Demographic Research a free, expedited, online journal
of peer-reviewed research and commentary
in the population sciences published by the
Max Planck Institute for Demographic Research
Konrad-Zuse Str. 1, D-18057 Rostock · GERMANY
www.demographic-research.org
DEMOGRAPHIC RESEARCH
VOLUME 12, ARTICLE 13, PAGES 323-380
PUBLISHED 28 JUNE 2005
www.demographic-research.org/Volumes/Vol12/13
DOI: 10.4054/DemRes.2005.12.13
Research Article
Geographical diversity of cause-of-death
patterns and trends in Russia
Jacques Vallin
Evgeni Andreev
France Meslé
Vladimir Shkolnikov
© 2005 Max-Planck-Gesellschaft.
Table of Contents
1
Introduction
324
2
2.1
2.2
2.2.1
2.2.2
Data and methods
Grouping causes of death
Hierarchical cluster analysis
Period-specific geographical patterns
Time-scale overall geographical patterns
335
335
336
338
339
3
339
3.1
3.2
3.3
3.4
To each period its own geographical pattern of
causes of death
1969-1970
1978-79
1988-89
1993-94
4
Constant geographical contrasts
350
5
Four main clusters explanatory of how
geographical contrasts contribute to general
mortality dynamics
Mortality dynamics by clusters
Contribution of geographical dynamics to allRussia mortality changes
358
6
Conclusion
365
7
Acknowledgements
367
References
368
Annex I
371
Annex II
374
Annex III
378
Annex IV
379
5.1
5.2
342
344
346
348
358
363
Demographic Research: Volume 12, Article 13
a research article
Geographical diversity of cause-of-death
patterns and trends in Russia
Jacques Vallin 1
Evgeni Andreev 2
France Meslé 3
Vladimir Shkolnikov 4
Abstract
This paper performs a systematic analysis of all currently available Russian data on
mortality by region, census year (1970, 1979, 1989, and 1994) and cause of death. It
investigates what links may be found between these geographical variations in causespecific mortality, the negative general trends observed since 1965, and the wide
fluctuations of the last two decades. For that, four two-year periods of observation were
selected where it was possible to calculate fairly reliable mortality indicators by
geographic units using census data for 1970, 1979, 1989, and micro-census data for
1994, and a clustering model was used.
Behind the complexity of the studied universe, three main conclusions appeared.
Firstly, in European Russia, there is a stark contrast between south-west and north-east,
both in terms of total mortality and of cause-of-death patterns. Secondly, analysis of
overall cause-of-death patterns for all periods combined clearly confirms that contrast at
the whole country level by the prolongation of the southern part of European Russia
through the continuation of the black soil (“chernoziom”) belt along the Kazakhstan
border, while the rest of Siberia presents a radically different picture to European
Russia. Thirdly, while it is difficult to infer any permanent geographical pattern of
mortality from that very fluctuating piece of history, 1988-89 appears to be a base
period for at least the entire period from 1969-1994.
1
Institut national d'études démographiques, E-mail: vallin@ined.fr
2
Centre of Demography and Human Ecology, Moscow, E-mail: evg_andreev@ns.cnt.ru
3
Institut national d'études démographiques, E-mail: mesle@ined.fr
4
Max Planck Institute for Demographic Research, E-mail: Shkolnikov@demogr.mpg.de
http://www.demographic-research.org
323
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
1. Introduction
Russian life expectancy has been stagnating for females and declining for males since
the mid-1960s, with a wide fluctuation in the 1980s and 90s related to the 1985 antialcohol campaign and the 1992-93 socio-economic crisis (Figure 1). The long-run
adverse trends are mainly due to increases in cardiovascular diseases, alcohol-related
mortality and violent deaths (Shkolnikov et al., 1996; Meslé et al., 2003). Changes in
alcohol consumption are responsible for the abrupt rise in life expectancy observed in
1985-86, and the subsequent decrease of 1990-92 (Meslé et al., 1994; Shkolnikov and
Nemtsov, 1997; Leon et al. 1997), while the even sharper decrease observed in 1992-94
stems from more varied causes of death, and is involved with social and economic
difficulties encountered by individuals and families in the transition to a market
economy (Meslé et al., 1998; Shkolnikov et al., 1998; Gavrilova et al., 2001). With
adaptation of people to the new economic situation, life expectancy returned to its
previous levels, but since 1998, it has resumed its long-term downwards trend (Meslé et
al., 2003).
It is, however, a matter of record that mortality varies within Russia from place to
place. Patterns of regional mortality variation have been documented in a number of
studies (Andreev, 1979; Shkolnikov, 1987; Shkolnikov and Vassin, 1994; Vassin and
Costello, 1997; Jozan and Prokhorskas, 1997; Walberg et al., 1998).
The general pattern of mortality increase from south-west to north-east, with lower
mortality in the blacksoil regions of southern European Russia and the southern part of
West Siberia, and higher mortality in northern European Russia, Ural, Siberia and the
far East, was identified by Evgueni Andreev (1979) from 1970 data, and by Vladimir
Shkolnikov (1987) from mortality data around the all-Soviet censuses of 1970 and
1979. The latter study found correlations between this geographical pattern and interregional differences in general socio-economic development, climate conditions, and
alcohol consumption. Studies by Vladimir Shkolnikov and Serguei Vassin (1994), and
by Peter Jozan and Remigijus Prokhorskas (1997), described geographical patterns of
mortality from principal and “avoidable” causes around the following census of 1989.
Serguei Vassin and Christine Costello (1997) used the same data to classify Russian
regions by the shapes of their mortality age curves. Finally, Peder Walberg et al. (1998)
analyzed decreases in male life expectancy from 1989 to 1994 (the last Russian microcensus) across regions of European Russia, and found some associations between these
decreases and the acuity of labor market changes.
This paper performs a systematic analysis of all currently available Russian data
on mortality by region, census year (1970, 1979, 1989, and 1994) and cause of death. It
investigates what links may be found between these geographical variations in causespecific mortality, the negative general trends observed since 1965, and the wide
324
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
fluctuations of the last two decades. For that, four two-year periods of observation were
selected where it was possible to calculate fairly reliable mortality indicators by
geographic units using census data for 1970, 1979, 1989, and micro-census data for
1994. As Figure 1 shows, 1970 may be taken as the base year (1965 would have been
preferable, but it is not a census year); 1979 is the closest census year prior to the antialcohol campaign; 1989 is fairly representative for the highest post-campaign life
expectancy, and 1994 is the lowest point at the peak of the socio-economic crisis.
Unfortunately, no reliable geographic mortality data are available for the most recent
years 5. However, these four points may already go a long way to explaining the
relationships between time changes in life expectancy and geographical variations of
mortality.
Also of interest would be to use life expectancy by regions as an indicator of
geographical variations of mortality. However, life expectancy has complicated nonlinear relations with cause-specific mortality depending on their weights and age
patterns. In particular, life expectancy is not easily decomposable by cause (Shkolnikov
et al., 2001). In this sense, age-standardized death rate (SDR) is a more suitable
indicator for a geographical analysis of cause-specific mortality. The choice of indicator
will have little impact on the outcome since the all-cause SDR is closely correlated with
life expectancy at birth across the Russian regions (Table 1).
Table 1:
5
Correlation between SDR and life expectancy at birth among
the 73 Russian administrative units
Sex
All periods
1969-70
1978-79
1988-89
1993-94
Males
-0.91
-0.88
-0.86
-0.75
-0.86
Females
-0.90
-0.89
-0.92
-0.87
-0.92
The 2002 census results are not yet available.
http://www.demographic-research.org
325
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
For each of the 73 administrative units used here 6 (see Annex I) SDR 7 were
computed on the basis of deaths by age and sex-specific mortality in the immediately
pre-census and census years. Since censuses were taken at the very beginning of the
census year 8, it was possible to use the age-specific population numbers reported by the
census as denominators.
Figure 2 shows the geographical variations in age-standardized mortality rates for
each of the four periods. For each sex, the four maps were drawn according to the same
7 value classes. The size of these classes is of one standard deviation, and the central
one is centred on the mean all-Russia value for the four periods. In fact, since the range
of values is much wider above than below this mean value, the central class is the third
one 9.
When considering the maps, it must be borne in mind that the Russian population
is very unequally distributed over its territory. A large part of the total population lives
in administrative units mostly concentrated in the European part of the country and the
south-western part of Siberia. By contrast, the very large units of northern Siberia,
especially Tyumen Oblast, Krasnoyarsk Kray, and the Republic of Sakha (Yakutia), are
sparsely-populated. For that reason, when interpreting the maps, more weight must be
given to colour variations among the smaller western and southern areas than the large
Siberian ones. Moreover, the population of territories like Tyumen Oblast or
Krasnoyarsk Kray, is mainly concentrated in their very small southernmost tips
(identified by dotted lines on the maps). Unfortunately, however, no data were available
for these administrative sub-units, and so the entire territory is coloured as for that small
part, which may give a misleading impression at first glance.
6
The present Russian territory is divided into 2 cities (Moscow and St Petersburg), 49 oblasts, 6 krays, 21 republics, 1 autonomous district
(Сhukchi AD), and 1 autonomous oblast (Jewish AO) (Goskomstat, 2002, p. 13). Prior to 1991, instead of these 21 republics, there were 15
“autonomous republics” that were direct members of the Russian Federation (including the Chechnya-Ingushia, which, in 1993, was split into
two republics: Chechnya and Ingushia), and 4 “autonomous oblasts” that were parts of 4 krays. Сhukchi autonomous district was part of
Magadan Oblast and the Jewish autonomous oblast was included in Khabarovsk Kray. Data for these 5 autonomous oblasts, 1 autonomous
district, and for Chechnya and Ingushia separately, were not available and we can only use here data at the level of whole krays and the whole
autonomous republic of Chechnya-Ingushia. Accordingly, 73 geographical units (2 cities, 49 oblasts, 6 krays, and 16 republics, see Annex I) are
used here. Furthermore, no data is available for Chechnya for 1993-94, and the entire autonomous republic of Chechnya-Ingushia has been left in
blank.
7
8
SDR were computed on the basis of the WHO (1992) standard European population.
Exact census dates were 15 January 1970, 17 January 1979, and 12 January 1989. In 1994, the micro-census accounted for 14 February;
however, this was not used directly as a denominator, but only to correct the post-census population estimate for 1 January 1994.
9
To depict those maps in life expectancy terms, mean life expectancies of each class of SDR are given in Annex IV.
326
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure1:
Trends in life expectancy since 1965 and the four points selected for
geographical analysis
Life expectancy
80
Females
75
1970
70
1979
1989
1994
65
60
1989
1970
55
50
1965
http://www.demographic-research.org
1979
Males
1975
1994
1985
1995
327
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Figure 2a:
Variation of age-standardized mortality rate among
73 administrative units in 1969-70, 1978-79,
1988-89 and 1993-94. Males
1969-70
1978-79
328
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure 2a:
Males (continued)
1989
1994
http://www.demographic-research.org
329
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Figure 2b:
Variation of age-standardized mortality rate among
73 administrative units in 1969-70, 1978-79,
1988-89 and 1993-94. Females
1969-70
1978-79
330
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure 2b:
Females (continued)
1988-89
1993-94
http://www.demographic-research.org
331
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
On the male side (Figure 2a), in 1970, a large part of Russia was less than half a
standard deviation from the mean (1970-1994) all-Russia SDR value. This included
most oblasts of central European Russia and almost all Siberian territories, and
accounted for 58.6% of the total male population (Table 2). This means that, at a time
when the big post-war gains in life expectancy from the decline in infectious diseases
had occurred, Russia had grown increasingly homogeneous. However, two main
deviations from the mean were observed: a much better situation in southern European
Russia plus Moscow (32.1% of the total population), and a slightly worse one in its
northern part (5.3%). The far East (Magadan, Kamchatka, Sakhalin) was affected by
much higher mortality, but, again, these are sparsely-populated territories (1.1%). Some
Caucasian republics (Dagestan, Kabardin-Balkar, North Ossetia, Chechnya-Ingushia),
accounting for 2.9% of the total population, seemed to enjoy notably low mortality, but
part at least of that advantage may only be apparent, stemming from under-registration
of infant and old-age mortality (Andreev and Kvasha, 2002; Meslé et al., 2003).
Table 2:
Percentage of total population living in regions included
in each of the SDR intervals, by sex.
Males
SDR intervals
1970
1979
1989
1994
<12.22
0.8
0
0
0
12.22-15.66
2.1
2.1
1.2
0
15.66 to 19.10
32.1
5.5
54.2
1.3
19.10 to 22.54
58.6
71.8
43.8
9.4
22.54 to 25.98
5.3
19.4
0.8
61.1
25.98 to 29.42
0.8
1.2
0
23.5
>= 29.42
0.3
0
0
3.9
Not available
0.9
Females
SDR intervals
1970
1979
1989
1994
<6.92
0.8
0.8
0
0
6.92-8.62
1.9
1.7
1.2
0
8.62 to 10.32
23.6
16.0
24.4
1.8
10.32 to 12.02
59
61.3
68.5
23.7
12.02 to 13.72
14.1
18.5
4.6
49.5
13.72 to 15.43
0.5
1.4
0.8
21.7
>= 15.43
0
0.3
0.5
2.5
Not available
332
0.9
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
In 1979, after ten years of rising mortality, the map has shifted towards higher
level classes of SDR and the yellow areas (71.8% of population) have expanded, while
the two blue ones have shrunk (5.5% and 2.1%). In particular, in European Russia, the
northern disadvantage has spread towards a large number of new oblasts, while the
southern advantage involves only a very small number of territories.
By contrast, in 1989, the remarkable progress made with the anti-alcohol campaign
has produced a reverse shift (Shkolnikov and Nemtsov, 1997). The low mortality area
expands extensively from southern European Russia to many oblasts along the border
between Siberia and Central Asia (again, bearing in mind that the large blue splash of
Tyumen Oblast is only due to the weight of its small southernmost part), and the two
blue areas account for 55.4% of the total male population. Moscow is again within that
advantaged area, now joined by St Petersburg. By contrast, only two special far-Eastern
territories (0.8%) are still affected by slightly higher mortality.
From 1989 to 1994, all territories experience a sharp rise in mortality. Almost the
entire map is shaded in high mortality colours, with only a couple of territories,
accounting for 10.7% of the total male population, remaining in the three lower
mortality classes (Table 2).
Apart from these shifts in value levels between successive maps, it can be
observed that the homogenisation at work in the 1960s continued in the 1970s. The
unweighted standard deviation for SDR was 2.95 for males in 1969-70, falling to 2.64
in 1978-79 (Table 3), with an accompanying, though less pronounced, decrease for the
weighted10 standard deviation. In other words, the decrease in life expectancy was
higher for advantaged than disadvantaged areas. By contrast, the gains due to the antialcohol campaign were far more marked for disadvantaged areas, and the unweighted
standard deviation fell dramatically to 1.70 in 1988-89, and even more so for the
weighted SD. Correspondingly, the socio-economic crisis had a much greater impact on
disadvantaged areas, while the unweighted standard deviation of SDR jumped to 3.01
in 1993-94 (2.32 for weighted SD).
10
As the unweighted standard deviation give the same weight to all regions regardless of population, it is useful to also check trends in standard
deviation weighted by population size.
http://www.demographic-research.org
333
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Table 3:
Standard deviation of SDR (unweighted and weighted
by population of regions) among 73 administrative units
at four periods
Sex
1969-70
1978-79
1988-89
1993-94
Unweighted
2.95
2.64
1.70
3.01
Weighted
2.42
2.36
1.32
2.32
Unweighted
1.43
1.51
1.65
1.46
Weighted
1.35
1.29
1.28
1.10
Males
Females
At the same time, the asymmetric geographical distribution mapped in Figure 2
minimises in 1988-89, while increasing in the other three periods11.
In all periods, the geography of female mortality is much more homogeneous than
that of males (Table 3). Furthermore, standard deviation (weighted or not) varies very
little from one period to the other, meaning that neither trends nor fluctuation greatly
affect mortality level distribution. It is true that trends and fluctuation are equally less
pronounced among females than males. Where male life expectancies declined, female
expectancies stagnated, and the wide fluctuations observed among males were much
slighter among females. At a lower level of contrast, however, women experienced the
same type of geographical variations as men in all periods.
How does the cause-of-death structure of mortality contribute to these
geographical changes over time? The first thing is to know if, for a specific period,
11
Asymmetry a can be measured by the ratio of the distance from the median M to the upper limit Q1 of the first quartile to that from M to the
upper limit Q3 of the third quartile:
a=
M − Q1
Q3 − M
Thus, for the SDR sets mapped in Figure 2, a takes the following values :
Sex
1969-70
1978-79
1988-89
1993-94
Males
0.82
0.95
1.07
0.91
Females
0.65
0.85
1.25
0.58
For both sexes, a is relatively closer to 1 in 1988-89, indicating minimum asymmetry that year. Homogeneisation is thus accompanied by a
decrease in asymmetry. Furthermore, 1988-89 is the only period where asymmetry is on the lower mortality side, which implies a reversal of
asymmetry.
334
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
geographical regularities may be found in the cause-of-death structure of mortality. It
would be then interesting to determine whether such regularities display any timebound continuity in spite of or in relation to trends and changes in overall mortality.
The first evident requirement is to select a way of measuring the distances between
geographical units, in terms of cause-of-death patterns. Since geographical and time
contrasts are more pronounced among males, the following analysis will be limited to
them.
2. Data and methods
Two questions must be addressed: How to use discontinuous cause–of-death statistics?
How to cluster geographical variations in cause-specific mortality?
2.1 Grouping causes of death
When using cause-of-death data at such a level, two problems are encountered. First,
observed time-bound changes are partly due to disruptions in statistical time series.
Second, observed geographical differences can be influenced by variations in data
quality.
Since the late sixties, several classifications of causes of death have been in use in
Russia, and deaths published according to old Soviet classifications had to be
reclassified accordingly. As a result of a previous work, all cause-of-death data have
been reclassified according to the most recent Soviet classification12 (1981, as modified
in 1988 for violent deaths) for all Russia, for each year, from 1956 to 1998 (Meslé et
al., 1996 and 2003). For this particular geographical analysis, all-Russia transition
coefficients13 have been applied to cause-of-death data by region for the two-year
period used14.
12
During the Soviet era, a specific cause-of-death classification was applied in all the Republics. Even after 1990, that classification remained in
use for around a decade in the new independent countries. The Soviet classification itself changed over time, and four successive versions were
used in the review period. Over time, however, its main structure evolved to more closely mirror the WHO International Classification of
Diseases, although it remains much less detailed (Meslé et al., 1996).
13
The method of reclassification used for the all-Russia level relies on transition coefficients from one classification to the next, calculated after a
complete analysis of medical definitions and statistical contents of individual items of both classifications (Meslé et al, 1996).
14
Assuming that changes in classification had the same impact in each region as at all-Russia level, is certainly not always true. However
computerised complete cause-of-death time series with which to check the actual differences are not available at the region level for the whole
period. The regional analysis, therefore, is based on an assumption that the differences will not be significant for the results.
http://www.demographic-research.org
335
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
After testing different possibilities, the 175 items of the Soviet classification have
been put into 16 groups meaningful for the regional analysis. That grouping retains as
such the causes of death that are responsible for high mortality rates, but also includes
more specific groups that could be clearly discriminant in geographic terms (Table 4).
In spite of the apparent specificity of their title, the items “atherosclerotic
cardiosclerosis” (with or without hypertension) were merged with “ill-defined deaths”
because, in the Soviet system still in use, “atherosclerotic cardiosclerosis” includes a
significant proportion of imprecise diagnoses that should have been classified as “illdefined” (Meslé and Vallin, 2003) (See also Section II).
Using such broad groups of causes partly but not entirely circumvents the data
quality issue. In the Soviet system, causes of death were identified and coded at the
local level and only the results were aggregated at higher levels. Theoretically, such a
system amply accommodates geographical variations in cause determination. However,
it may also be argued that the powerful Soviet era bureaucracy left local administration
very little room for manoeuvre. Unfortunately, there are no specific studies of
geographical variations in cause-of-death data quality that we could ascertain, such that
we must use the available data "as found", with the caveat that part of our observations
will be coloured by that issue.
2.2 Hierarchical cluster analysis
We have therefore to work with a 3-dimensional dataset for 73 regions, 4 periods and
16 cause-of-death groups. Most standard methods assume more or less symmetric and
interchangeable categories, which is not the case here, since our purpose is to analyse
geographical mortality and its dynamics over 4 periods. A straightforward analysis of
the 3-dimensional matrix is arguably inappropriate, as interactions among the three
dimensions would interfere with interpretation of the results. What is known about
cause-specific mortality trends in Russia from the 1970s-1990s suggests that time-scale
and regional variations may be greater for certain causes of death, and that these causes
may differ between the regional and time-scale dimensions (Meslé et al., 2003;
Shkolnikov and Vassin, 1994; Vassin and Costello, 1997). To gain insight into these
peculiarities, regions were first clustered by their cause-of-death profiles independently
for each time-period, after which an aggregated clustering was performed based on 64dimensional (16x4) vectors, which made it possible to identify a more general structure
characteristic for the whole time-period on average. Hierarchical analysis was applied
to all clusters.
336
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
That analysis, based on Euclidian distance between cause-specific SDR patterns15
of the 73 geographical units (Annex I), was applied using within-group linkage
methods16. This procedure attempts to identify relatively homogeneous clusters of
geographical units based on selected characteristics, using an algorithm that starts with
each unit in a separate cluster and combines clusters until only one is left. Thus, the
numbers of clusters change in the process of calculation.
Table 4:
Group
number
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
15
Groups of causes used for geographical patterns
of mortality by causes of death
Causes of death
Infectious diseases
Stomach cancer
Other digestive cancers
Cancer of larynx, trachea, bronchus and lung
Other cancers
Atherosclerotic cardiosclerosis + Ill-defined
Ischaemic heart diseases
Other heart diseases
Cerebrovascular diseases and other diseases of the
circulatory system
Influenza and pneumonia
Chronic respiratory diseases
Digestive diseases
Other diseases
Accidental alcohol poisoning
Suicide, homicide, and injury undetermined whether
accidentally or purposely inflicted
Other external causes
Soviet Classification
item numbers
1-44
47
45, 46, 48-51
52, 53
54-67
92, 93, 158, 159
90, 91, 94, 95
84-89, 96, 97
98-102
104-107
103, 108-114,
115-127,
68-83, 128-157
163
173-175
160-162, 164-172
In fact, Russian mortality trends are heavily influenced by age-group-specific problems, especially at adult ages. A forthcoming study will focus
on changes in the geographical mortality patterns in adult males only. This study seeks only to estimate changes in total mortality.
16
SPSS program, release 11.5 (SPSS, 2002), was used for the calculations.
http://www.demographic-research.org
337
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
2.2.1 Period-specific geographical patterns
For our classification, each territory in period t was presented as a vector of 16 values
W(r,t,i):
W (r , t , i) =
SDR(r , t , i )
,
σ SDR (t , i)
where:
SDR is the standardized death rate,
r is the geographical unit (r = 0 for Russia),
t represents the period (1969-1970, 1978-1979; 1988-1989, or 1993-1994),
i represents the cause-of-death group (from 1 to 16),
σ SDR (t , i ) is the standard deviation among geographical units for the cause i during the
period t.
To obtain the clusters for a specific period, the classification was made on vectors
of the 16 values of W of each unit, and the distance between units is the Euclidian
distance weighted by the Russian level of mortality for each cause, weighted by the
square root of the all-Russia SDR ( SDR (0, t , i) ), or:
(
)
1
2 2

 ∑ (W (r2 , t , i ) − W (r1 , t , i )) ⋅ SDR(0, t , i ) 
 i

Weighting is used here to reduce the distance's sensitivity to fluctuations in
mortality from causes with relatively low, but highly variable, mortality. The SDRweightings are square-rooted to reduce the cluster analysis's dependency on cause–ofdeath choices (see Annex III). The distance function with such weightings would be
unaffected by, for example, the division of a broad cause into two more specific causes.
The distances thus calculated for each pair of units were then compared to produce
clusters. Each unit starts out as a cluster. In step two, the two closest units are
aggregated into one cluster, and the distances between clusters are recalculated. In step
three, the two closest clusters are again merged, and so on until a reasonable number of
clusters is reached. The analysis was continued until all units were aggregated into just
two clusters, from which point the previous results were compared to the last one to
determine the materiality of each additional cluster. Finally, a division into 10 clusters
appeared to be most informative.
338
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
2.2.2 Time-scale overall geographical patterns
To analyse the persistence of geographical cause-of-death patterns over time, a global
clustering for all four periods together was obtained on the basis of the distances
between the vectors of 16 causes multiplied by 4 periods, i. e. 16x4=64 values of W for
each geographical unit. Once again, Euclidian distances are weighted by
SDR (0, t , i) :
(
)
1
2 2

 ∑∑ (W (r2 , t , i ) − W (r1 , t , i )) ⋅ SDR(0, t , i ) 
 t i

3. To each period its own geographical pattern of causes of death
Figure 3 shows the results of the cluster analysis applied to each of the four periods,
independently. Although such maps are not comparable, and must be interpreted
independently of one another, we defined colour sets for each to maximize ease of
interpretation. The results of the time-scale global analysis (next section) were used for
that, and detailed explanations are given in annex II. The principle is to assign the same
colour to the period-specific clusters that are least distant from the four-period cluster as
that assigned in the global analysis. For example, in the four following maps, areas of
the cluster that was closest to the first cluster of the global analysis were shaded red,
and so on. The same colours were therefore used for clusters showing similarities. In
some cases, however, the same global cluster was selected as the closest for two periodspecific clusters of the same period. In these cases, the same colour was used, but with
distinctive hatching. In this way, for example, the individual cluster of Kamchatka
Oblast is associated in 1988-89 with global cluster number 6, already associated with St
Petersburg and Moscow cities that formed another cluster in that period. So, while each
map shows the results for 10 clusters, the colour sets of some maps include a specific
11th colour, while one of the ten basic colours is missing (see Table II-2 in Annex II).
In all periods, each Figure 3 map shows dissimilarities that echo some of those
appearing in Figure 2. In particular, the northern and southern parts of Western Russia
present contrasting pictures. However, these contrasts are less evident, as the links
between cause-of-death patterns and total mortality levels are complex. There are also
wide between-period variations in cause-of-death geography, requiring each period to
be considered independently. To clarify the differences between clusters, stars have
been plotted for each according to the standardized mortality rates (SMR) by cause of
http://www.demographic-research.org
339
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Figure 3:
Males. Geographical distribution of 10 clusters according
to cause-of-death structure
1969-70
99
8
101
Murmans k Oblas t
3
Kali ningr adO blast
10
St. P eter sburg
Republ ci of Kare lia
11
13
Leni ngrad Obl ast
12
Magadan Obl ast andC hukchi Autonom ous O kru g
P sko v Obl ast
24
5
Novgor od Obla st
25
7
T ver Obl ast
S mol ensk Obl ast
15
21
18
Y arosl avl Oblas t
20
Br yansk O blast
Vol ogda Oblast
27
17
16
26
22
4
19
M oscow
K aluga Obl ast
M oscow Oblast
37
R epubli c of Kom i
Kam chatk a Obl ast
Kostr oma Obl ast
Iv anovo Obl ast
33
V a
l dim ir O blast
Tul aO blast
Or yol Oblast
23
38
Kurs k Obl ast
35
32
36
30
39
Kir ov Obla st
29
Ni h
z n y Novgor odO blast
31
83
45
Republ ci of Mor dovia
Tam bov Oblast
62
Chuvash Republ ci
44
65
P erm O blast
42
48
Penza Obl ast
Udm urt r epubl ci
T yumen Obl ast
67
U ly anovsk Obla st
59
Repu blic of Tat ars t an
46
47
61
Sar atov Obl ast
Volgogr ad Obl ast
S amar a Oblast
R ostov Obl ast
64
58
68
Republ ci of Bashkor tost an
Republ ci of Kalm ykia
43
95
63
41
Chely abi nsk O blast
53
55
75
74
Kur ganO blast
Or enbur gO blast
Astr akhan Oblast
Kr asnoyars k Kray and Republi c of K hakasia
S verdl ovsk Obl ast
Kr asnodar kray andR epubli c of A dyg eya
S tavr opol Kr ay and Kar achaev-C ir cass i anr epubl ci
97
Re public of Sakha (Y akuti a)
R epubli c of M ari y E l
Vo ronezh Obl ast
57
91
76
Ryazan Obl ast
Li petz k Obl ast
Bel gorod Obl ast
Khabar ovsk K ray andJew is hA utonom ous O blast
56
O msk O blast
100
86
Tom sk O blast
K abardi an- Balk ar r epubl ci
S akhal in Obl ast
73
Republ ci of Nor th Osset ia
96
C hechenand Ing ush Republ ci s
Ir kut sk Obl ast
51
Republ ci of Dagestan
N ovosibi sr k Obla st
72
88
Am ur Obl ast
80
1
K emer ov O blast
Chit aO blast
71
2
Republ ic of Bur yat i a
94
Al tai Kr ay a ndRe public of Al tai
81
Republ ic of Tuva
3
4
P ri mor sky kr ai
5
St. Petersburg
6
7
8
9
Moscow
10
No Data
1978-79
99
8
101
Murmansk Oblas t
3
Kali ningr adO blast
10
St. P eter sbur g
Republ ic of Kar eli a
11
13
Leni ngra dOb last
12
M agadan Obl ast andCh ukchi A utonom ous O krug
Pskov O blast
24
5
No vgorod Obl ast
25
7
Tver Obl ast
S mol ensk O blast
15
21
18
20
Br yansk O blast
37
V ologda Obl ast
27
M oscow Obl ast
26
19
17
16
Republ ic of Komi
Kam chatk a Obl ast
Kostr om aO blast
I van ovoO blast
Tu la Oblast
38
V ladim ir Obl ast
33
30
Ni zhny Novgor od Obla st
23
91
32
36
39
T ambov Obl ast
V oronezh Obl ast
Ki rov Obl ast
29
31
83
Republ ic of Mor dovia
44
62
Chuvash Republ ic
42
48
Udm urt republ ic
Tyum enO blast
67
Uly anovsk Obl ast
R epubli c of Tat ars t an
59
61
Sar atov Obl ast
Sam ara Obl ast
64
68
R epubli c of Bashkor tost an
Republ ic of Kalm ykia
43
Astr akhan Oblast
Or enbur gO blast
95
63
41
Chel yabinsk Oblast
53
55
K rasnoyar sk Kr ay andRepu blic of Khakasi a
Sverd lovsk O blast
46
47
Volg ograd Obl ast
R ostov Obl ast
58
65
Per m Obl ast
Kr asnodar kray andR epubli c of A dygeya
Stavr opol Kr ay and Kar achaev-C ir cassian e
r publ ic
Republ ic of Sakha( Yakut ia)
Republ ic of M ar iy El
45
Pen za Obl ast
97
76
Ryazan Obl ast
Lipet zk Obl ast
Bel gorod Obl ast
57
4
Yar oslavl Obl ast
M oscow
K aluga Obl ast
22
Or yol Obl ast
Kur sk Obl ast
35
Kur ganO blast
75
74
Khabar ovsk Kr ay an dJewi shAu tonom ous O blast
Tomsk Obl ast
56
Omsk Obl ast
100
86
Kabar dian- Balk ar r epubl ic
Rep ublic of Nor th Osset ia
Sakhal in Obl ast
73
96
C hechenand n
I gush Repub lic s
I r kuts k O blast
51
Republ ic of Dagesta n
Novosi bir sk Obl ast
72
88
A mur O blast
80
1
2
Kem erov Obl ast
C hit aO blast
71
Republi c of B ury at ia
A lt ai Kray andR epubli c of A lta i
81
94
Republi c of Tuva
3
4
5
P ri mor sky kr ai
St. Petersburg
6
7
Moscow
8
9
10
No Data
340
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure 3:
(Continued)
1988-89
99
8
101
Murmansk Oblas t
3
Kali ningr adO blast
10
St. P eter sbur g
Republ ic of Kar eli a
11
13
Leni ngra dOb last
12
M agadan Obl ast andCh ukchi A utonom ous O krug
Pskov O blast
24
5
No vgorod Obl ast
25
7
Tver Obl ast
S mol ensk O blast
15
21
18
M oscow Obl ast
26
17
16
Republ ic of Komi
Kam chatk a Obl ast
Kostr om aO blast
I van ovoO blast
Tu la Oblast
Or yol Obl ast
V ladim ir Obl ast
33
30
Ni zhny Novgor od Obla st
23
38
Kur sk Obl ast
35
91
32
39
36
Ki rov Obl ast
29
31
83
Republ ic of Mor dovia
62
Chuvash Republ ic
65
Per m Obl ast
42
48
Pen za Obl ast
44
Udm urt republ ic
Tyum enO blast
67
Uly anovsk Obl ast
R epubli c of Tat ars t an
59
61
Sam ara Obl ast
64
58
68
R epubli c of Bashkor tost an
Republ ic of Kalm ykia
43
Astr akhan Oblast
95
63
41
Chel yabinsk Oblast
53
55
K rasnoyar sk Kr ay andRepu blci of Khakasi a
Sverd lovsk O blast
46
47
Sar atov Obl ast
Volg ograd Obl ast
R ostov Obl ast
Kr asnodar kray andR epubli c of A dygeya
Stavr opol Kr ay and Kar achaev-C ir cassian republ ic
Republ ic of Sakha( Yakut ia)
Republ ic of M ar iy El
45
T ambov Obl ast
V oronezh Obl ast
97
76
Ryazan Obl ast
Lipet zk Obl ast
Bel gorod Obl ast
57
19
M oscow
K aluga Obl ast
22
4
Yar oslavl Obl ast
20
Br yansk O blast
37
V ologda Obl ast
27
75
74
Kur ganO blast
Or enbur gO blast
Khabar ovsk Kr ay an dJewi shAu tonom ous O blast
Tomsk Obl ast
56
Omsk Obl ast
100
86
Kabar dian- Balk ar r epubl ic
Sakhal in Obl ast
73
Rep ublic of Nor th Osset ia
96
C hechenand In gush Repub lic s
I r kuts k O blast
51
Republ ic of Dagesta n
Novosi bir sk Obl ast
72
88
A mur O blast
80
1
Kem erov Obl ast
C hit aO blast
71
2
Republi c of B ury at ia
94
A lt ai Kray andR epubli c of A lta i
81
Republi c of Tuva
3
4
P ri mor sky kr ai
5
St. Petersburg
6
7
Moscow
8
9
10
No Data
1993-94
99
8
101
Murmansk Oblas t
3
Kali ningr adO blast
10
St. P eter sbur g
Republ ic of Kar eli a
11
13
Leni ngra dOb last
12
M agadan Obl ast andCh ukchi A utonom ous O krug
Pskov O blast
24
5
No vgorod Obl ast
25
7
Tver Obl ast
S mol ensk O blast
15
21
18
Br yansk O blast
V ologda Obl ast
20
27
26
22
4
Yar oslavl Obl ast
19
M oscow
K aluga Obl ast
M oscow Obl ast
37
17
16
Republ ic of Komi
Kam chatk a Obl ast
Kostr om aO blast
I van ovoO blast
33
V ladim ir Obl ast
Tu la Oblast
Or yol Obl ast
38
23
32
91
97
Kur sk Obl ast
35
76
Ryazan Obl ast
Lipet zk Obl ast
39
Bel gorod Obl ast
36
30
Ki rov Obl ast
Ni zhny Novgor od Obla st
29
31
83
Republ ic of Sakha( Yakut ia)
Republ ic of M ar iy El
T ambov Obl ast
45
Republ ic of Mor dovia
62
Chuvash Republ ic
V oronezh Obl ast
65
Per m Obl ast
Pen za Obl ast
44
42
48
Udm urt republ ic
Tyum enO blast
67
Uly anovsk Obl ast
R epubli c of Tat ars t an
59
57
61
Sar atov Obl ast
Sam ara Obl ast
64
58
68
95
63
R epubli c of Bashkor tost an
41
Chel yabinsk Oblast
Kur ganO blast
Stavr opol Kr ay and Kar achaev-C ir cassian republ ic
53
55
Republ ic of Kalm ykia
K rasnoyar sk Kr ay andRepu blci of Khakasi a
Sverd lovsk O blast
46
47
Volg ograd Obl ast
R ostov Obl ast
Kr asnodar kray andR epubli c of A dygeya
43
Astr akhan Oblast
75
74
Khabar ovsk Kr ay an dJewi shAu tonom ous O blast
Or enbur gO blast
Tomsk Obl ast
56
Omsk Obl ast
100
86
Kabar dian- Balk ar r epubl ic
Rep ublic of Nor th Osset ia
Sakhal in Obl ast
73
96
C hechenand In gush Repub lic s
I r kuts k O blast
51
Republ ic of Dagesta n
Novosi bir sk Obl ast
72
88
A mur O blast
80
1
2
Kem erov Obl ast
C hit aO blast
71
Republi c of B ury at ia
A lt ai Kray andR epubli c of A lta i
81
94
Republi c of Tuva
3
4
5
P ri mor sky kr ai
St. Petersburg
6
7
8
9
Moscow
10
No Data
http://www.demographic-research.org
341
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
death for the cluster, divided by the Russian SMR. The stars are shown by period in
Figures 4a to 4d. For each period, two graphs were plotted. The first aggregates stars of
the four major clusters (those closest to the first 4 global clusters). The second
aggregates the six remaining clusters, which are mostly quite singular clusters with very
specific cause-of-death patterns, but comprising only a few territorial units. Most
significance attaches to the shape of the first four stars.
3.1 1969-1970
Notwithstanding that Russian male life expectancy had been declining since 1965, the
situation found at the end of the sixties still largely reflects the outcomes of previous
decades' strong gains from the country-wide reduction of infectious diseases that
produced greater homogeneity. This largely explains why most territorial entities fall in
the same cluster (n° 2), coloured light pink on the 1969-70 map. This cluster aggregates
40 of the 73 units, including Moscow and St Petersburg (represented by two circles
below the map to delineate them more clearly). The star for this cluster shown at the top
of figure 4a appears clearly very closed to circle 1 representing the all-Russia cause-ofdeath pattern. The second most important cluster (cluster 3, in dark blue on the map)
aggregates 10 units, all located in the northern European part of Russia. It is also very
close to the Russian mean, except for two groups of causes: stomach cancer and
cerebrovascular diseases17. Stomach cancer was also a cause of excess mortality in a
third cluster geographically located in the same part of Russia (cluster 4, in light blue),
however was not particularly affected by cerebrovascular diseases, but showed strong
excess mortality from atherosclerotic cardiosclerosis. The latter group is known to be
quite ill-defined, and probably includes many circulatory system diseases, including
cerebrovascular disease (Meslé et al., 1998), although that fact may not be significant.
By contrast, excess mortality from accidental alcohol poisoning, suicide and homicide
is more representative of that cluster. Both of these groups are closely associated with
alcohol consumption in Russia (Shkolnikov and Nemtsov, 1997). Alcohol consumption
and especially accidental alcohol poisoning, is an even greater cause of excess mortality
in the cluster shaded red (cluster 1), which aggregates 10 units, also mainly located in
the northern part of European Russia, plus 3 units in western Siberia. That cluster is
also strongly characterised by chronic respiratory diseases and to a lesser extent by
17
That group also includes "other diseases of the circulatory system", but that component is much less significant than cerebrovascular diseases,
so the entire group is labelled “cerebrovascular diseases” here for simplicity's sake.
342
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure 4a:
Cause-of-death patterns of each cluster compared to the all-Russia
cluster (cluster SMR/Russian SMR). Ten clusters, males, 1969-70
Other external causes
Infectious diseases
1.8
#1
Cancer of stomach
1.6
uicide + Homicide+Injury undetermined w hether
accidentally or purposely inflicted
#2
1.4
Other digestive cancers
1.2
#3
#4
1
0.8
Accidental poisoning by alcohol
Cancer of larynx, trachea, bronchus and lung
0.6
0.4
0.2
Other diseases
Other cancers
0
Digestive diseases
Atherosclerotic cardiosclerosis + Ill defined
Chronic respiratory diseases
Other ischaemic heart diseases
Influenza+pneumonia
Other heart diseases
Cerebrovascular diseases + Other diseases of the
circulatory system
Other external causes
Infectious diseases
4
#5
#6
Cancer of stomach
#7
3.5
Suicide + Homicide+Injury undetermined w hether
accidentally or purposely inflicted
3
Other digestive cancers
#11
2
Accidental poisoning by alcohol
#9
#10
2.5
Cancer of larynx, trachea, bronchus and lung
1.5
1
0.5
0
Other diseases
Other cancers
Digestive diseases
Atherosclerotic cardiosclerosis + Ill defined
Chronic respiratory diseases
Other ischaemic heart diseases
Influenza+pneumonia
Other heart diseases
Cerebrovascular diseases + Other diseases of the
circulatory system
http://www.demographic-research.org
343
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
infectious diseases. There is also a notable excess mortality from atherosclerotic
cardiosclerosis, but once again, this category is probably misleading, since its
importance is very likely associated with an underestimation of cardiovascular mortality
(and especially of “other ischaemic heart diseases”).
Of the other clusters, 5 comprise a single territorial unit. Each appears to display
exceptional excess mortality for one specific group of causes. For example, in Yakutia
(cluster 10), "other digestive cancers" mortality was 4 times higher than in Russia in
that period, while chronic respiratory diseases are more than 3.5 times higher in the
Mari Republic (cluster 11) than for all-Russia. Likewise, in Magadan (cluster 6),
digestive diseases mortality was 4 times that of Russia, while influenza and pneumonia
mortality was 4 times higher in the republic of Tuva (cluster 9). More interestingly, a
specific cluster (n° 7) aggregating 4 units that comprise a continuous zone in the
Caucasian region (the Republics of Dagestan, Chechnya-Ingouchia, North-Ossetia and
Kabardin-Balkaria), is characterised by very low mortality for almost all causes except
“infectious diseases” and “influenza and pneumonia”. Its “accidental alcohol
poisoning” mortality was almost negligible and its “suicide and homicide” mortality
was only one third that of Russia.
3.2 1978-79
The difference between 1969-70 and 1978-79 portrays here the results of the first step
of a long-term decline in life expectancy. In 1978-79, cause-of-death patterns appear
more heterogeneous among territories. The very large cluster coloured light pink on the
1969-70 map is split into two clusters coloured light pink (cluster 2) and red (cluster 1)
on the 1978-79 map. While the light pink cluster still presents very similar cause-ofdeath pattern, quite close to the Russian mean, the red one is now different, but also
close to the Russian mean (Figure 4b). They are distinguished by the fact that their
slight differences from the all-Russia pattern are not the same (alcohol-related mortality
is higher in the light pink cluster and lower in the red one, while the converse is true for
“other ischaemic heart diseases” and cerebrovascular diseases).
By contrast, the new dark blue cluster (n° 3) tends to aggregate territories in the
former dark and light blue clusters (northern part of European Russia), but is in fact
characterised by a cause-of-death pattern closer to that of the former light blue than
dark blue one. Indeed, it is mainly associated with excess mortality in alcohol-related
causes of death and atherosclerotic cardiosclerosis, and to a lesser extent, stomach
cancer. The fourth most interesting cluster in that period is the one comprised of the two
big Russian cities (cluster 6, in dark green on the 1978-79 map). At that time, these
344
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure 4b:
Cause-of-death patterns of each cluster compared to the all-Russia
cluster (cluster SMR/Russian SMR). Ten clusters, males, 1978-79
Other external causes
Infectious diseases
1,6
#1
Cancer of stomach
1,4
uicide + Homicide+Injury undetermined w hether
accidentally or purposely inflicted
#2
1,2
#3
Other digestive cancers
#6
1
0,8
Accidental poisoning by alcohol
Cancer of larynx, trachea, bronchus and lung
0,6
0,4
0,2
0
Other diseases
Other cancers
Digestive diseases
Atherosclerotic cardiosclerosis + Ill defined
Chronic respiratory diseases
Other ischaemic heart diseases
Influenza+pneumonia
Other heart diseases
Cerebrovascular diseases + Other diseases of the
circulatory system
Other external causes
Infectious diseases
4,5
#4
#5
Cancer of stomach
4
#7
3,5
uicide + Homicide+Injury undetermined w hether
accidentally or purposely inflicted
Other digestive cancers
3
#11
2
Accidental poisoning by alcohol
#9
#10
2,5
Cancer of larynx, trachea, bronchus and lung
1,5
1
0,5
0
Other diseases
Other cancers
Digestive diseases
Atherosclerotic cardiosclerosis + Ill defined
Chronic respiratory diseases
Other ischaemic heart diseases
Influenza+pneumonia
Other heart diseases
Cerebrovascular diseases + Other diseases of the
circulatory system
http://www.demographic-research.org
345
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
cities showed a specific cause-of-death pattern with higher mortality from all cancers
(especially digestive cancer), “other ischaemic heart diseases” and “other diseases”. To
a large extent, this pattern typically reflects improved diagnostic accuracy. By contrast,
violent deaths, infectious diseases, and chronic respiratory diseases are much less
significant than for all-Russia.
Once again, the lower part of Figure 4b displays stars for marginal clusters formed
by small numbers of units. The five single-unit clusters (7 Magadan, 4 Kamchatka,
9 Tuva, 10 Yakutia, and 11 Dagestan) show marked peculiarities, almost identical to
1969-70 when the same unit is concerned. By contrast, the remaining cluster, grey-blue
in 1978-79, corresponds to a part of the red cluster of 1969-70, aggregating the
Republics of Mari, Chuvach, and Udmurtia, and Kirov Oblast (cluster 5). That cluster
effectively shows a cause-of-death pattern close to that of the red cluster of 1969-70.
3.3 1988-89
The period 1988-89 characterised the situation produced by the positive results of the
Gorbachev anti-alcoholism campaign. At that time, as shown in table 2, the
geographical heterogeneity of total mortality was much less than during the other three
review periods. Life expectancy gains from reductions in alcohol- and cardiovascular
diseases-related mortality had been higher in regions where mortality was high. The
1988-89 map in Figure 3 also shows a much clearer geographical structure of cause of
deaths than in any other period. There is a very sharp contrast between the northern and
southern parts of European Russia, and it is the only time where there is a clear
demarcation between the European part (here extended to the southern tip of western
Siberia) and the Asian part of Russia. Homogenisation in term of levels attributable to
the reduction in alcohol-related causes of death, which have a spurious effect on causeof-death patterns, seems to have been associated with the emergence of a geographical
pattern more related to the more fundamental cause-of-death structure.
The cause-of-death patterns of the three main clusters, coloured light pink
(southern European Russia, cluster 2), light blue (northern European Russia, cluster 4),
and dark blue (most of Siberia, cluster 3), are quite close to the all-Russia pattern. They
are slightly discriminated, however, by specific causes of death: alcohol-related causes
are more significant in northern European Russia and much less in southern European
Russia, but the converse is true for infectious diseases. Most cardiovascular categories
and infectious diseases are more significant in Siberia.
A fourth cluster (n° 1), coloured red, aggregates a series of geographic units along
a West-East axis from the Baltic Sea (Kaliningrad Oblast) through Kirov Oblast,
346
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure 4c:
Cause-of-death patterns of each cluster compared to the all-Russia
cluster (cluster SMR/Russian SMR). Ten clusters, males, 1988-89
Other external causes
Infectious diseases
1,6
#1
Cancer of stomach
#2
1,4
uicide + Homicide+Injury undetermined w hether
accidentally or purposely inflicted
#3
1,2
Other digestive cancers
#4
1
#11
0,8
Accidental poisoning by alcohol
Cancer of larynx, trachea, bronchus and lung
0,6
0,4
0,2
0
Other diseases
Other cancers
Digestive diseases
Atherosclerotic cardiosclerosis + Ill defined
Chronic respiratory diseases
Other ischaemic heart diseases
Influenza+pneumonia
Other heart diseases
Cerebrovascular diseases + Other diseases of the
circulatory system
Other external causes
Infectious diseases
4,5
#6
Cancer of stomach
#7
4
#8
3,5
uicide + Homicide+Injury undetermined w hether
accidentally or purposely inflicted
Other digestive cancers
#9
3
#10
2,5
2
Accidental poisoning by alcohol
Cancer of larynx, trachea, bronchus and lung
1,5
1
0,5
0
Other diseases
Other cancers
Digestive diseases
Atherosclerotic cardiosclerosis + Ill defined
Chronic respiratory diseases
Other ischaemic heart diseases
Influenza+pneumonia
Other heart diseases
Cerebrovascular diseases + Other diseases of the
circulatory system
http://www.demographic-research.org
347
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Udmurt Republic, Bashkortostan Republic, and Kurgan Oblast to the Altaï. Departing
appreciably from the Russian pattern, it is marked by higher mortality from chronic
respiratory diseases, atherotic cardio-sclerosis and accidental alcohol poisoning.
As in the previous period, Moscow and St Petersburg form specific cluster n° 11.
Exceptionally, this is included with the four main ones to account for the population
sizes of these two entities. Also, that cluster's profile departs much less from the mean
than the following ones, included in the second star graph. The 1988-89 Moscow and St
Petersburg profile closely matches that of the previous period. However, mortality from
“other ischaemic heart diseases” is below the Russian mean, compared to above it in the
previous period, and the converse applies to the broad category of atherosclerotic
cardiosclerosis and ill-defined causes. In fact, this difference from the previous period
reflects less any real change in the disease profile than in coding practice. This is
because a new instruction from the Health Ministry in 1989 required any death over the
age of 80 where no specific cause was mentioned in the medical file or autopsy report
to be classed as "senility", and forbade deaths at younger ages from being reported as
acute cardiovascular failure unless the diagnosis was confirmed by an autopsy report
(Meslé and Vallin, 2003). The first requirement resulted mainly in transfers from
atherosclerotic cardiosclerosis to ill-defined causes, which is without consequence for
the broad group used here, but the second led to numerous transfers of specific
cardiovascular diseases (here aggregated under “other ischaemic heart diseases”)
toward ill-defined causes, which is the main explanatory factor in the change of the star
for that cluster.
As usual, small clusters generally comprising a single geographic unit, present
singular profiles very far from the Russian mean. Significantly, however, the second
graph of Figure 4b shows Yakutia no longer identified as a specific cluster, but
incorporated in the large cluster 3.
3.4 1993-94
The fourth period is dominated by the effects of the economic and social crisis.
Unsurprisingly, the corresponding map (Figure 3) shows clusters differently demarcated
and much less compact than the previous three. The northern part of European Russia,
very homogeneous and specific in the previous period, is now divided into four
different clusters, each also including what may be far-flung other regions. What is
shown by the Figure 4d stars has very little in common with what could be said of the
previous ones. The most salient feature is that total mortality increased dramatically,
with specific impacts in terms of causes of death, giving new geographical forms to the
348
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure 4d:
Cause-of-death patterns of each cluster compared to the all-Russia
cluster (cluster SMR/Russian SMR). Ten clusters, males, 1993-94
Other external causes
Infectious diseases
1,8
#1
Cancer of stomach
1,6
#2
1,4
uicide + Homicide+Injury undetermined w hether
accidentally or purposely inflicted
#3
Other digestive cancers
1,2
#4
1
0,8
Accidental poisoning by alcohol
Cancer of larynx, trachea, bronchus and lung
0,6
0,4
0,2
0
Other diseases
Other cancers
Digestive diseases
Atherosclerotic cardiosclerosis + Ill defined
Chronic respiratory diseases
Other ischaemic heart diseases
Influenza+pneumonia
Other heart diseases
Cerebrovascular diseases + Other diseases of the
circulatory system
Other external causes
Infectious diseases
4
#5
#6
Cancer of stomach
3,5
uicide + Homicide+Injury undetermined w hether
accidentally or purposely inflicted
#7
3
Other digestive cancers
#10
2
Accidental poisoning by alcohol
#8
#9
2,5
Cancer of larynx, trachea, bronchus and lung
1,5
1
0,5
0
Other diseases
Other cancers
Digestive diseases
Atherosclerotic cardiosclerosis + Ill defined
Chronic respiratory diseases
Other ischaemic heart diseases
Influenza+pneumonia
Other heart diseases
Cerebrovascular diseases + Other diseases of the
circulatory system
http://www.demographic-research.org
349
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
clusters and new disease profiles to their stars. One general effect of that change is to
make the disease profiles closer to the mean and less different from each other
(Figure 4d). The first cluster (in red) that includes for the first time Moscow and St
Petersburg is close to the Russian mean for all causes of death except atheroclerotic
cardiosclerosis and ill-defined causes. It aggregates fairly well developed regions,
including the two metropolitan areas where the mortality rise closely mirrored the
national mean, but where also the new coding rules mentioned above were still applied.
The second cluster (in pink), showing the most favourable profile, with often submean mortalities, relatively low impact of accidental alcohol poisoning and greater
incidence of other heart diseases, has been reduced to a much smaller number of
geographic units, almost entirely limited to Caucasian regions. For that reason, it is also
marked by a relative importance of digestive and infectious diseases. Clusters 3 (dark
blue) and 4 (light blue) display considerable geographical discontinuity, both including
regions in the far East as well as middle Siberia and north-western European Russia.
Mortality levels are almost invariably equal to or above the mean for all causes. The
main difference between the two lies in “other ischaemic heart diseases”, which are the
cause of much higher mortality in the third than fourth cluster.
Smaller clusters are more singular than ever. Kamchatka (cluster 5), Magadan
(cluster 7), Tuva (cluster 10), and Dagestan (cluster 8), in particular, are once more
isolated in specific clusters with generally the same type of deviations from the mean.
For Kamchatka, however, the impact of accidental alcohol poisoning is very prominent.
4. Constant geographical contrasts
The wide mortality fluctuations and general time trends notwithstanding, are there any
major geographical features that can be reliably retained when aggregating data for all
four periods? To characterize the global geographical pattern, we used the distance
measurement between the vectors of 16 causes multiplied by four periods for each
geographical unit, as explained earlier in section I-B-2. As previously, we first defined a
set of 73 clusters (as many clusters as geographical units), incrementally merging the
two closest clusters, leaving only two clusters at the end, splitting all-Russia into two
geographical entities.
This final result, shown in the first chart of Figure 5, reveals only an approximate
mix of geographical differences, but is already highly illustrative of the geographical
structure of Russian mortality. One interesting point is that the units aggregated in these
two geographical entities display a high degree of geographical continuity. This
includes mainly areas of the western and southern parts of European Russia, extended
by a belt of south-western Siberia at the Kazakhstan border. The inclusion of Chita
350
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
oblast and the Republic of Sakha (former Yakoutia) arguably slightly weakens that
continuity, but the importance of these very large sparsely-populated areas should
probably not be over-stated. These two units aside, the second cluster clearly aggregates
the rest of Asian Russia with all the north-eastern part of European Russia. A second
point is that these two final clusters embrace two very significant and contrasting parts
of geographical and historical Russia: the rich agricultural part running from the
Northern Caucasus to the Lithuanian border, the Southern Volga basin and the Siberian
prolongation of the chernoziom (black soil) belt, and the less hospitable regions of
northern European Russia and Siberia. All-period mortality has been consistently lower
in the former than the latter part. And Figure 6 shows that these two big clusters present
two very different disease profiles.
Mortality is lower for almost all causes in cluster 1 than in cluster 2. The only
exception is for chronic respiratory diseases that produce a slightly higher mortality in
cluster 1 than all-Russia. However, the difference is much less than for the high excess
mortality revealed by cluster 2 for two large groups of causes: violent deaths (including
alcohol-related) and cardiovascular mortality. The standardized mortality rate by
accidental alcohol poisoning is 40% above the Russian mean in cluster 2, and slightly
below in cluster 1. Excess mortality in cluster 2 is also highly significant for the two
other groups of violent death, especially suicide and homicide. Likewise, that cluster
presents high excess mortality for the three broad groups of cardio-vascular diseases
(ischaemic heart diseases, other heart diseases and cerebrovascular diseases). Another
major cause of excess mortality for that group is related to acute respiratory diseases
(influenza and pneumonia). Even digestive and respiratory cancer mortalities are also
above the Russian mean.
However, there is a useful purpose to looking beyond this approximate split into
the two largest clusters, and refining the analysis by returning to the previous steps of
our clustering procedure. The three remaining charts of Figure 5 show the maps with 4,
6, and 8 clusters, and, finally, Figure 7, the map with 10 clusters. Arguably, the latter is
the most appropriate for final comments. This is because the 4- and 6-cluster maps only
isolate very specific geographical units, without disrupting the big areas shown by the
2-cluster map. It is only with the 8-cluster map that the first large area is split into two
large parts, while the 10-cluster map (Figure 7) does likewise for the second large area.
And these two large partitions are, again, of great geographical interest.
Of the 10 clusters, six comprise a single (clusters 10 Yakoutia, 6 Magadan,
5 Kamchatka, 9 Tuva, and 7 Dagestan) or at most two (cluster 8: Republic of Mordovia
and Chuvach Republic) geographic units. These units are the same as those frequently
isolated by the single period analyses.
http://www.demographic-research.org
351
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Figure 5:
Global clustering for the four periods together,
according to 2 to 8 clusters. Males
Males, 2 clusters
99
8
101
Murmans k Oblas t
3
Kali ningr adO blast
10
Republ ci of Kare lia
St. P eter sburg
11
13
Leni ngrad Obl ast
12
Magadan Obl ast andC hukchi Autonom ous O kru g
P sko v Obl ast
24
5
Novgor od Obla st
25
7
T ver Obl ast
S mol ensk Obl ast
15
27
M oscow Oblast
26
17
16
R epubli c of Kom i
Kam chatk a Obl ast
Kostr oma Obl ast
V a
l dim ir O blast
33
30
Ni h
z n y Novgor odO blast
Iv anovo Obl ast
Tul aO blast
Or yol Oblast
23
38
Kurs k Obl ast
35
91
32
39
36
Kir ov Obla st
29
31
83
Republ ci of Mor dovia
62
Chuvash Republ ci
44
65
P erm O blast
42
48
Penza Obl ast
Udm urt r epubl ci
T yumen Obl ast
67
U ly anovsk Obla st
Repu blic of Tat ars t an
59
46
47
61
Sar atov Obl ast
Volgogr ad Obl ast
S amar a Oblast
R ostov Obl ast
64
58
68
Republ ci of Bashkor tost an
Republ ci of Kalm ykia
43
95
63
41
Chely abi nsk O blast
53
55
75
74
Kur ganO blast
Or enbur gO blast
Astr akhan Oblast
Kr asnoyars k Kray and Republi c of K hakasia
S verdl ovsk Obl ast
Kr asnodar kray andR epubli c of A dyg eya
S tavr opol Kr ay and Kar achaev-C ir cass i anr epubl ci
Re public of Sakha (Y akuti a)
R epubli c of M ari y E l
45
Tam bov Oblast
Vo ronezh Obl ast
97
76
Ryazan Obl ast
Li petz k Obl ast
Bel gorod Obl ast
57
19
M oscow
K aluga Obl ast
22
4
Y arosl avl Oblas t
20
Br yansk O blast
37
Vol ogda Oblast
21
18
Khabar ovsk K ray andJew is hA utonom ous O blast
56
O msk O blast
100
86
Tom sk O blast
K abardi an- Balk ar r epubl ci
S akhal in Obl ast
73
Republ ci of Nor th Osset ia
96
C hechenand Ing ush Republ ci s
Ir kut sk Obl ast
51
Republ ci of Dagestan
N ovosibi sr k Obla st
72
88
Am ur Obl ast
80
1
K emer ov O blast
Chit aO blast
71
Republ ic of Bur yat i a
2
94
Al tai Kr ay a ndRe public of Al tai
81
No Data
Republ ic of Tuva
P ri mor sky kr ai
St. Petersburg
Moscow
Males, 4 clusters
99
8
101
Murmans k Oblas t
3
Kali ningr adO blast
10
Republ ci of Kare lia
St. P eter sburg
11
13
Leni ngrad Obl ast
12
Magadan Obl ast andC hukchi Autonom ous O kru g
P sko v Obl ast
24
5
Novgor od Obla st
25
7
T ver Obl ast
S mol ensk Obl ast
15
20
Br yansk O blast
Vol ogda Oblast
21
18
27
19
K aluga Obl ast
M oscow Oblast
37
16
26
22
4
Y arosl avl Oblas t
M oscow
17
R epubli c of Kom i
Kam chatk a Obl ast
Kostr oma Obl ast
33
Iv anovo Obl ast
V a
l dim ir O blast
Tul aO blast
Or yol Oblast
23
38
91
32
97
Kurs k Obl ast
35
76
Ryazan Obl ast
Li petz k Obl ast
39
Bel gorod Obl ast
36
Tam bov Oblast
30
Kir ov Obla st
Ni h
z n y Novgor odO blast
29
31
83
Republ ci of Mor dovia
Penza Obl ast
44
62
Chuvash Republ ci
65
P erm O blast
42
48
Udm urt r epubl ci
T yumen Obl ast
67
U ly anovsk Obla st
Repu blic of Tat ars t an
59
57
47
Volgogr ad Obl ast
61
S amar a Oblast
64
58
68
Republ ci of Bashkor tost an
Chely abi nsk O blast
53
55
43
Or enbur gO blast
Astr akhan Oblast
95
63
41
Republ ci of Kalm ykia
Kr asnoyars k Kray and Republi c of K hakasia
S verdl ovsk Obl ast
46
Sar atov Obl ast
R ostov Obl ast
Kr asnodar kray andR epubli c of A dyg eya
S tavr opol Kr ay and Kar achaev-C ir cass i anr epubl ci
Re public of Sakha (Y akuti a)
R epubli c of M ari y E l
45
Vo ronezh Obl ast
Kur ganO blast
75
74
Khabar ovsk K ray andJew is hA utonom ous O blast
56
O msk O blast
Republ ci of Nor th Osset ia
100
86
Tom sk O blast
K abardi an- Balk ar r epubl ci
S akhal in Obl ast
73
96
C hechenand Ing ush Republ ci s
Ir kut sk Obl ast
51
Republ ci of Dagestan
N ovosibi sr k Obla st
72
88
Am ur Obl ast
80
1
K emer ov O blast
Chit aO blast
71
Republ ic of Bur yat i a
2
Al tai Kr ay a ndRe public of Al tai
3
81
94
Republ ic of Tuva
P ri mor sky kr ai
4
No Data
St. Petersburg
Moscow
352
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure 5:
(Continued)
Males, 6 clusters
99
Murmansk Oblas t
8
101
3
Kali ningr adO blast
10
St. P eter sbur g
Republ ic of Kar eli a
11
13
Leni ngra dOb last
12
M agadan Obl ast andCh ukchi A utonom ous O krug
Pskov O blast
24
5
No vgorod Obl ast
25
7
Tver Obl ast
S mol ensk O blast
15
21
18
Br yansk O blast
V ologda Obl ast
27
20
Yar oslavl Obl ast
K aluga Obl ast
M oscow Obl ast
17
16
26
22
37
Republ ic of Komi
Kam chatk a Obl ast
Kostr om aO blast
V ladim ir Obl ast
33
30
Ni zhny Novgor od Obla st
I van ovoO blast
Tu la Oblast
Or yol Obl ast
23
38
Kur sk Obl ast
35
91
32
36
39
Ki rov Obl ast
29
31
83
Republ ic of Mor dovia
62
Chuvash Republ ic
65
Per m Obl ast
42
48
Pen za Obl ast
44
Udm urt republ ic
Tyum enO blast
67
Uly anovsk Obl ast
R epubli c of Tat ars t an
59
47
61
Sam ara Obl ast
68
64
95
R epubli c of Bashkor tost an
63
41
58
Chel yabinsk Oblast
53
55
43
Republ ic of Kalm ykia
K rasnoyar sk Kr ay andRepu blci of Khakasi a
Sverd lovsk O blast
46
Sar atov Obl ast
Volg ograd Obl ast
R ostov Obl ast
Kr asnodar kray andR epubli c of A dygeya
Stavr opol Kr ay and Kar achaev-C ir cassian republ ic
Republ ic of Sakha( Yakut ia)
Republ ic of M ar iy El
45
T ambov Obl ast
V oronezh Obl ast
97
76
Ryazan Obl ast
Lipet zk Obl ast
Bel gorod Obl ast
57
4
19
M oscow
75
74
Kur ganO blast
Or enbur gO blast
Astr akhan Oblast
Khabar ovsk Kr ay an dJewi shAu tonom ous O blast
Tomsk Obl ast
56
Omsk Obl ast
100
86
Kabar dian- Balk ar r epubl ic
Rep ublic of Nor th Osset ia
Sakhal in Obl ast
73
96
C hechenand In gush Repub lic s
I r kuts k O blast
51
Republ ic of Dagesta n
Novosi bir sk Obl ast
72
88
A mur O blast
80
1
Kem erov Obl ast
C hit aO blast
71
Republi c of B ury at ia
2
94
A lt ai Kray andR epubli c of A lta i
81
3
Republi c of Tuva
P ri mor sky kr ai
4
5
St. Petersburg
6
No Data
Moscow
Males, 8 clusters
99
8
101
Murmansk Oblas t
3
Kali ningr adO blast
10
Republ ic of Kar eli a
St. P eter sbur g
11
13
Leni ngra dOb last
12
M agadan Obl ast andCh ukchi A utonom ous O krug
Pskov O blast
24
5
No vgorod Obl ast
25
7
Tver Obl ast
S mol ensk O blast
15
20
Br yansk O blast
37
V ologda Obl ast
21
18
27
M oscow Obl ast
26
16
17
Republ ic of Komi
Kam chatk a Obl ast
Kostr om aO blast
V ladim ir Obl ast
33
30
Ni zhny Novgor od Obla st
I van ovoO blast
Tu la Oblast
Or yol Obl ast
38
Kur sk Obl ast
35
23
91
32
36
39
T ambov Obl ast
V oronezh Obl ast
Ki o
r v Obl ast
29
31
83
45
44
62
Chuvash Republ ic
Udm urt republ ic
Tyum enO blast
67
Uly anovsk Obl ast
R epubli c of Tat ars t an
59
47
Volg ograd Obl ast
61
Sam ara Obl ast
64
68
R epubli c of Bashkor tost an
Republ ic of Kalm ykia
43
Or enbur gO blast
Astr akhan Oblast
95
63
41
Chel yabinsk Oblast
53
55
K rasnoyar sk Kr ay andRepu blic of Khakasi a
Sverd lovsk O blast
46
Sar atov Obl ast
R ostov Obl ast
58
65
Per m Obl ast
42
48
Kr asnodar kray andR epubli c of A dygeya
Stavr opol Kr ay and Kar achaev-C ir cassian republ ic
Republ ic of Sakha( Yakut ia)
Republ ic of M ar iy El
Republ ic of Mor dovia
Pen za Obl ast
97
76
Ryazan Obl ast
Lipet zk Obl ast
Bel gorod Obl ast
57
19
M oscow
K aluga Obl ast
22
4
Yar oslavl Obl ast
Kur ganO blast
75
74
Khabar ovsk Kr ay an dJewi shAu tonom ous O blast
56
Omsk Obl ast
100
86
Tomsk Obl ast
Kabar dian- Balk ar r epubl ic
Rep ublic of Nor h
t Osset ia
Sakhal in Obl ast
73
96
C hechenand In gush Repub lic s
I r kuts k O blast
51
Republ ic of Dagesta n
Novosi bir sk Obl ast
72
88
A mur O blast
80
1
Kem erov Obl ast
C hit aO blast
71
Republi c of B ury at ia
2
A lt ai Kray andR epubli c of A lta i
3
81
94
Republi c of Tuva
P ri mor sky kr ai
4
5
St. Petersburg
6
7
Moscow
8
No Data
http://www.demographic-research.org
353
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Figure 6:
Cause-of-death patterns of the two final clusters compared
to the Russian cluster (cluster SMR/Russian SMR).
Males, four periods altogether
Other external causes
Infec tious diseases
1,6
Cancer of stomach
1,4
uicide + Homicide+Injury undetermined w hether
accidentally or purposely inf licted
#1
1,2
#2
Other digestive cancers
1
0,8
Accidental poisoning by alc ohol
Cancer of larynx, trachea, bronchus and lung
0,6
0,4
0,2
Other diseas es
Other cancers
0
Digestive diseases
Atherosclerotic cardioscleros is + Ill defined
Chronic respiratory diseases
Other ischaemic heart diseases
Influenza+pneumonia
Other heart diseases
Cerebrovascular diseases + Other diseas es of the
circulatory system
Figure 7:
Global clustering for the four periods combined,
according to 10 clusters. Males
99
8
101
Murmansk Oblast
3
K ali nin gra dO bla st
10
St . Pet er sbur g
R epubl ic of K are li a
11
13
Leni ngr ad Obl ast
12
P skov Obl ast
24
M agad anO bl ast and Ch ukchi Au tono mou sO kru g
5
N ovgor od Obl ast
25
7
Tver O bl ast
Sm ol ensk Obl ast
15
Vol ogda O blast
21
18
20
27
Yar osl avl O blast
M oscow O blast
16
26
22
4
19
Mo scow
Kal uga O blast
Br yansk Obl ast
37
17
R epubl ic of Ko mi
K am chat kaO bl ast
Ko str om aO bl ast
Iva novo Obl ast
V ladi m ir O bl ast
Tul aO bl ast
Or yol O bl ast
38
Kur sk Obl ast
35
23
33
B elg oro dO bla st
36
39
Tam bov Obl ast
30
29
31
83
Rep ubli cof Mo rdovi a
62
Chu vash Repub lic
44
65
P er m O blast
42
48
Ud mur t r epub lic
Tyum en Ob last
67
U lyano vskO bla st
R epubl ic of T atar st an
59
47
V olgog a
r d O blast
61
S am ara O bla st
64
68
Re publ ic of Bas hkort ost an
95
63
41
Che lyabi nsk Obl ast
Kur gan O blas t
Sa
t vr opol K r ayan dK ar achaev- Ci r cassian r epubl ic
53
55
R epubl ic of K alm yki a
43
O re nbur g Obl ast
Ast r akhan O blast
K rasn oyar skK ray and R epubl ic of Kh akasia
Sver dl ovsk Obl ast
46
Sar at ovO bl ast
Ros tov Ob last
K r asnodar kr ay and Re publ ic of Ad ygeya
58
Re publ ic of Sak ha( Ya kuti a)
R epubl ic of M ari y El
45
P enza Obl ast
97
76
Kir ov O blast
Ni zhny Novg oro dO bl ast
V or onezh Ob last
57
91
32
Ryaza nO bl ast
Li pet zkO bla st
75
74
Khab aro vskK ra yand Jew ish Au tono mou sO bla st
Tom sk Obl ast
56
O msk Ob last
100
86
Kaba d
r i an- Bal kar r epu bli c
R epubl ic of N ort h Osse ti a
S akhali n Obl ast
73
96
C hechen and n
I gush Repu bli cs
I rkut sk Obl ast
51
R epubl ic of D agest an
No vosibi r skO bla st
72
88
A m ur O blast
80
1
Kem er ovO bl ast
Ch it aO bl ast
71
R epubl ic of Bu ryat ia
2
A lt ai Kr ay and Re publi co f Al a
ti
3
81
94
R epubl ic of Tuva
P ri mo rsky kr ai
4
5
St. Petersburg
6
7
Moscow
8
9
10
No Data
354
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
The other four clusters, by contrast, comprise numerous units, which, again,
present strong continuities. Two show a very clear split between the European
(cluster 4) and Asian (cluster 3) parts of the second cluster of the 2-cluster map,
adhering closely to the Ural frontier. The disjunction is less clear for the subdivision in
the first one, but roughly-speaking, it counterposes the southern (cluster 2) and central
(cluster 1) parts of European Russia, even if Moscow and St Petersburg are associated
with the southern part.
These 4 main clusters therefore concentrate a very large part of the total Russian
population. Figure 8 distributes the male Russian population (mean population for the
entire interval between the first and last period) among the 10 clusters, ranked by SDR
level (four-period mean). While the lowest and highest SDR levels are observed in
some extreme small clusters, the four main clusters vary closer to the mean. They differ
more by disease profile than overall mortality level. Figure 8 illustrates very clearly
their dominant demographic weight.
Figure 8:
Distribution of male Russian population by the 10 clusters,
ranked by level of SDR
http://www.demographic-research.org
355
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
The disease profiles, illustrated by the stars in Figure 9, show that the excess
mortality from chronic respiratory diseases observed in the final cluster 1 relates only to
the central part of European Russia. The northern Caucasus and the units along the
Kazakhstan border are not concerned at all. Conversely, the lower mortality from
accidental alcohol poisoning characterizes only southern, not central European Russia,
where it is slightly above the Russian mean. For these two disease groups, therefore, the
South is much better placed than central European Russia. The reverse is true for
cancers, which cause higher mortality in the South than in central European Russia,
which probably explains Moscow and St Petersburg's association to that cluster.
The difference between the northern part of European Russia and most of Siberia
is related to both overall mortality levels and disease profile. For most causes of death,
mortality is higher in the eastern than western Ural. But there are also marked
differences in terms of specific causes of death. First, its lower general mortality
notwithstanding, the European side presents higher mortalities for accidental alcohol
poisoning and stomach cancer. Second, mortality is particularly high in the Asian part
for violent deaths (suicide, homicide and other accidental deaths), infectious diseases
and acute respiratory diseases. And finally, mortality from heart diseases (ischaemic
and others) is also much higher in the Asian than the European part.
The second part of Figure 9 confirms what can be conjectured from period
analyses: clusters comprised of one or two geographic units only are highly singular,
with sharply contrasting disease profiles. Often, just one or two causes of death will
give a very clearly-defined star. What is less easy is to distinguish what is due to true
differences from what stems from insufficiently robust data. Continuity here stems
more from the individual units concerned being almost invariably the same than from
being same-cause related. To comment further would require a forensic examination of
each particular context, which is beyond the scope of this paper.
356
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure 9.
Cause-of-death patterns of each global cluster compared to the allRussia cluster (cluster SMR/Russian SMR). Ten clusters, males
Other external causes
Inf ectious diseases
1,4
#1
Cancer of stomach
1,2
uicide + Homicide+Injury undetermined w hether
accidentally or purposely inf licted
#2
#3
Other digestive cancers
1
#4
0,8
0,6
Accidental poisoning by alcohol
Cancer of larynx, trachea, bronchus and lung
0,4
0,2
Other diseases
0
Other cancers
Digestive diseases
Atherosclerotic cardiosclerosis + Ill def ined
Chronic respiratory diseases
Other ischaemic heart diseases
Influenza+pneumonia
Other heart diseases
Cerebrovascular diseases + Other diseases of the
circulatory system
Other external causes
Inf ectious diseases
3,5
#5
#6
Cancer of stomach
3
uicide + Homicide+Injury undetermined w hether
accidentally or purposely inflicted
#7
Other digestive cancers
2,5
#8
#9
2
#10
1,5
Accidental poisoning by alcohol
Cancer of larynx, trachea, bronchus and lung
1
0,5
Other diseases
0
Other cancers
Digestive diseases
Atherosclerotic cardiosclerosis + Ill def ined
Chronic respiratory diseases
Other ischaemic heart diseases
Inf luenza+pneumonia
Other heart diseases
Cerebrovascular diseases + Other diseases of the
circulatory system
http://www.demographic-research.org
357
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
5. Four main clusters explanatory of how geographical contrasts
contribute to general mortality dynamics
To determine the extent to which geographical contrasts contribute to general Russian
mortality dynamics, we focused on the four main clusters identified in the previous
section. For simplicity, these clusters are named as:
cluster 1 (red): Chernoziom,
cluster 2 (pink): South,
cluster 3 (dark blue): Siberia,
cluster 4 (light blue): European North.
5.1 Mortality dynamics by clusters
For each of these four clusters, we calculated the SDR for the 16 groups of causes at
each period. After merging various groups with strong similarities, Table 5 gives the
results for 12 new groups.
At the level of all-cause mortality, the dynamics of the four clusters are quite
similar (Figure 10). Differences between clusters are mostly a matter of level: the
Chernoziom cluster shows a trajectory that is almost identical to that for all-Russia. The
trajectories of the other three clusters are almost the same, but at a lower level for the
South cluster and a higher level for Siberia and the European North18. Significantly,
however, with the 1988-89 mortality decline due to the anti-alcohol campaign, the
differences between clusters narrowed. In particular, the European North and South
approached the all-Russia level, at a time when the European North and Siberia were
almost at the same level. The inter-cluster distance also increased somewhat from 1989
to 1994, as the mortality increase was slightly steeper for clusters 3 and 4 than for
clusters 1 and 2.
18
Only one cluster is below the all-Russia level, while three are equal to or above it. This is due to the fact that South, above the mean, is more
heavily populated than the other three clusters.
358
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Table 5:
Cluster
Male SDR (per 1000) by cause of death for the 4 main clusters
and 4 periods
Groups of causes
Total
1+10
2
3-5
6
7
8
9
11
12+13
14
15
16
1969-70
1
19.28
1.11
1.01
1.61
4.91
0.65
0.54
3.35
2.64
1.01
0.34
0.73
1.37
2
18.56
0.96
1.14
2.24
3.97
1.14
0.56
3.80
1.54
1.14
0.21
0.66
1.21
3
21.11
1.30
1.17
2.14
4.41
0.99
0.87
4.23
1.61
1.18
0.36
1.11
1.74
4
21.71
0.95
1.33
2.02
5.56
0.98
0.71
4.67
1.86
1.02
0.33
0.94
1.35
Russia
19.33
1.01
1.12
1.99
4.47
0.98
0.63
3.82
1.83
1.11
0.28
0.78
1.32
1
20.97
0.73
0.76
1.83
5.29
1.38
0.55
3.88
2.48
1.04
0.47
0.93
1.64
2
19.84
0.65
0.83
2.35
4.19
1.99
0.42
4.18
1.58
1.14
0.26
0.83
1.43
3
22.39
0.95
0.91
2.35
3.55
2.31
0.79
5.10
1.56
1.15
0.52
1.37
1.83
4
23.18
0.62
0.99
2.25
5.78
1.91
0.48
5.18
1.69
1.06
0.53
1.08
1.61
Russia
20.75
0.70
0.86
2.20
4.60
1.79
0.53
4.30
1.77
1.14
0.39
0.96
1.52
1
18.95
0.38
0.63
2.45
4.40
1.98
0.50
3.75
1.67
1.03
0.18
0.73
1.25
2
18.42
0.37
0.65
2.83
3.44
2.18
0.42
4.32
1.12
1.14
0.12
0.74
1.09
3
19.88
0.52
0.71
2.83
2.36
2.88
0.73
5.14
1.15
1.17
0.14
1.04
1.22
4
20.13
0.36
0.81
2.82
3.47
2.77
0.40
5.18
1.15
1.01
0.19
0.81
1.16
Russia
18.79
0.39
0.68
2.75
3.53
2.16
0.50
4.41
1.21
1.12
0.14
0.79
1.12
1
24.20
0.57
0.58
2.69
5.62
2.71
0.80
4.38
1.77
1.34
0.55
1.46
1.72
2
23.50
0.65
0.59
2.94
4.34
3.00
0.77
4.90
1.24
1.53
0.41
1.56
1.56
3
26.59
0.84
0.65
2.93
2.95
4.29
1.19
5.94
1.36
1.60
0.66
2.31
1.86
4
26.98
0.65
0.70
3.00
4.87
3.51
0.94
6.19
1.33
1.38
0.82
1.83
1.76
Russia
24.45
0.66
0.61
2.87
4.47
3.04
0.91
5.17
1.34
1.49
0.54
1.72
1.64
1978-79
1988-89
1993-94
Groups of causes:
1+10: Infectious diseases and influenza and pneumonia
2: Stomach cancer
3+4+5: Other cancers
6: Atherosclerotic cardiosclerosis + Ill-defined
7: Ischaemic heart diseases
8: Other heart diseases
9: Cerebrovascular diseases and other diseases of the circulatory system
11: Chronic respiratory diseases
12+13 Other diseases
14: Accidental alcohol poisoning
15: Suicide, homicide, and injury unspecified whether accidentally or purposely
inflicted
16: Other external causes
http://www.demographic-research.org
Clusters
1: (Red) Chernoziom
2: (Pink) South
3: (Dark blue) Siberia
4: (Light blue) European North
359
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Figure 10:
Change in male all-cause SDR in the 4 main clusters,
compared to Russia
SDR (per 1000)
29
27
25
23
3 Siberia
4 European North
21
19
2 South
1 Chernoziom
17
Russia
15
1969
1974
1979
1984
1989
1994
Almost the same phenomenon is observable here as when Russia is compared to
other European republics of the former USSR, like Ukraine or the Baltic countries
(Meslé and Shkolnikov, 2000; Meslé and Vallin, 2003; Meslé, 2004): so powerful and
generalized were the impacts of first Gorbachev's anti-alcohol campaign, and then the
socio-economic crisis due to the abrupt transition to a market economy, as to dominate
all the European Republics of the former USSR and all the main Russian regions in
comparable fashion. Only slight differences can be identified around these very
dramatic changes in general mortality dynamics.
However, greater differences appear when causes of death are taken into account
(Figure 11). In fact, causes of death add three types of differentiation that do not appear
at the level of all-cause mortality. First, some causes do not at all follow the big changes
that are typical of the general dynamic. Second, some clusters are more heavily
influenced than others by specific groups of causes. And, finally, some clusters show
very specific mortality schedules for some groups of causes.
360
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure 11:
Changes in SDR by 12 groups of causes in the 4 main clusters,
compared to Russia
SDR (per 1000)
SDR (per 1000)
SDR (per 1000)
3,0
3,0
3,0
2,5
1+10. Infectious diseases,
influenza & pneumonia
2,0
1,5
1,0
1,0
0,5
3
2
4
0,0
1969
1
1979
1989
0,5
2,5
2,0
4
2,0
1,5
3
2. Cancer of stomach
2,5
1
2,5
SDR (per 1000)
3,0
2,0
1,5
2
4
1,0
1
0,0
1969
1979
1989
1
1,5
3
3
11.Chronic respiratory
diseases
1,0
2
0,5
0,5
0,0
1969
0,0
1969
1979
1989
3+4+5. Other cancers
1979
SDR (per 1000)
SDR (per 1000)
SDR (per 1000)
SDR (per 1000)
6,5
6,5
6,5
6,5
7. Ischaemic heart
diseases
4
5,5
4
2
1989
8. Other heart diseases
5,5
5,5
5,5
4
1
4,5
4,5
3
4,5
4,5
2
2
3
3,5
1
4
3,5
3,5
3,5
2,5
2,5
1,5
1,5
3
2,5
2,5
2
1,5
6. Arteriosclerotic
cardiosclerosis & ill
defined causes
0,5
1969
1979
1989
1
1,5
3
4
0,5
1969
1979
1989
0,5
1969
1
1979
1989
9. Cerebro-vascular
and other circulatory
diseases
0,5
1969
1979
SDR (per 1000)
SDR (per 1000)
SDR (per 1000)
SDR (per 1000)
3,0
3,0
3,0
3,0
2,5
12+13. Other diseases
2,0
1,5
3
1 2
2,5
14. Accidental poisoning
by alcohol
2,5
2,0
2,0
1,5
1,5
15. Suicide, homicide,
etc.
2,0
1,5
4
1,0
1,0
0,5
0,5
0,0
1969
0,0
1969
1979
4
1
4
1989
0,5
2
1979
1989
1
1,0
0,0
1969
2
1979
16. Other external
causes
2,5
3
1,0
1989
3
4
1
2
0,5
1989
0,0
1969
1979
1989
1= Chernoziom
2=South
3 = Siberia
4 = European North
http://www.demographic-research.org
361
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
In the first category, cancers appear highly specific, in that changes are wholly
independent of the general scheme, either as steady, across-the-board decreases
(stomach cancers) or increases (other cancers) over the review period. The trends in
infectious diseases (including influenza and pneumonia), “other heart diseases” and all
“other diseases” present no major inter-cluster differentiation, and are quite independent
from the anti-alcohol campaign but markedly affected by the 1993-94 crisis.
Several notable examples can be seen of clusters demonstrating particular allperiod effects of specific groups of causes. The Chernoziom cluster, for instance, which
is the closest to the all-Russia level of all-cause mortality (Figure 10), shows much
higher chronic respiratory diseases mortality than the all-Russia level in all four
periods, and that mortality is continuously slightly below the mean in all the other three
clusters. By contrast, the same Chernoziom cluster is consistently well below Russia for
other cancers, while very slight differences are to be found between the other clusters
and all-Russia. A third geographical peculiarity of that type affects the Siberia cluster,
where the suicide and homicide SDR is constantly well above the rest of the country
including the European North cluster, the total mortality of which was higher on Figure
10.
Finally, different mortality schedules are shown for specific groups of causes. For
example, while South and Chernoziom clusters track the general schedule for
“Atherosclerotic cardiosclerosis & ill-defined causes” quite closely, the European North
and Siberia change quite differently. In the European North, mortality decline during
the anti-alcohol campaign is much more spectacular than elsewhere. What is
remarkable for Siberia is that mortality from that group of causes decreased
exceptionally between 1969-70 and 1979-80 while it was increasing everywhere else.
Some of these abnormalities may be explained by reverse observations made about
ischaemic heart diseases, however. In Siberia, mortality from that group of causes rose
far more than elsewhere in the first time interval and continued to increase rapidly in
the second, while the rate of increase slowed in the South and Chernoziom. Meanwhile,
in the European North that mortality has been rising steadily in both intervals.
However, the ischaemic heart disease-related SDR was significantly lower at the time
than that for the previous group of causes, especially in the European North. It is very
probable that in that part of Russia, a large proportion of mortality from the large illdefined category including “atherosclerotic cardiosclerosis” is alcohol-related, and that
the anti-alcohol campaign had a greater impact in that cluster than elsewhere. The
Siberia cluster's situation is probably more related to a general improvement in cause of
death registration, which resulted in a general reduction in the proportion of ill-defined
causes over the whole period.
The Siberia cluster also presents an atypical schedule, where the SDR from “other
external causes” was much higher than elsewhere in 1969-70, but much closer to the
362
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
mean in 1988-89 and in 1994-95. It may be that this part of Russia was more affected
by these causes during the rapid industrialization years, returning to the mean thereafter.
The main point here is to clarify the impact of the two big events that affected
mortality trends in the past 30 years: the anti-alcohol campaign and the socio-economic
crisis. Three main features can be singled out. First is the highly differential response of
causes of death. For some (cancers), the trends appeared quite independent of both
events, while others (infectious diseases, other diseases, other heart diseases) are
affected by the socio-economic crisis only. Yet others (violent deaths, most circulatory
diseases) are dramatically affected by both events.
Those causes of death whose trends are less affected by these events are also those
that display less geographical variation. This is particularly so for stomach cancer. By
contrast, geographical variations are wider for circulatory diseases and violent deaths,
and for the latter in particular, geographical distance increases as mortality rises and
decreases as mortality declines. More specifically, the 1993-94 socio-economic crisis
exacerbated geographical inequalities in terms of suicide, homicide and other external
causes.
However, differentiated trends, for most causes of death notwithstanding, the
hierarchy between the four clusters remained substantially unchanged, producing a
clearly constant inter-cluster hierarchy at the level of total mortality, as shown in
Figure 10.
5.2 Contribution of geographical dynamics to all-Russia mortality changes
Is there a discernible relation between the geographical clustering of mortality by cause
of death, and the observed changes in Russian mortality? We sought to answer that final
question by decomposing changes in the Russian SDR observed over the three time
intervals studied into the specific contributions of each cluster. Table 6 presents the
results for the 10 clusters in terms of total mortality. From 1969-70 to 1978-79
Russian19 male SDR increased by 142 per 100000. Then, from 1978-79 to 1988-89, it
decreased by 199, and finally, from 1988-89 to 1993-94, it increased again by 594.
Prima facie, a very large share of these changes was always attributable to the 4 first
clusters alone. This is mainly due to the fact that these 4 clusters concentrate the
overwhelming majority of the total mean population (95.6%). It can also be seen that
cluster 2, which accounts for 41% of the total population, consistently dominates the
all-Russia changes: it accounts for more than 30% of the total change at each period,
and even close to 40% in the first and third periods. But it is also clear that the three
19
In fact, only 72 regions are considered here as data for Chechnya are still not available.
http://www.demographic-research.org
363
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
other main clusters play different roles from period to period, despite having population
weights of the same magnitude. For example, cluster 1 accounts for 25% of the Russian
mortality increase in the first period, cluster 3 only 15% and cluster 4 under 20%. By
contrast, cluster 4 accounts for 26% of the mortality decrease in the second period,
while clusters 1 and 3 account for less than 20% each. So the first period deterioration
was more attributable to lower than higher mortality regions, while, conversely, higher
mortality regions contributed more to the mortality decline during the anti-alcohol
campaign. Finally, no such difference appeared during the third period of the huge
mortality increase due to the socio-economic crisis, which affected all Russian territory
more or less equally.
Table 6:
Contribution of each of the 10 clusters to the changes in
Russian SDR for three periods
Absolute changes in SDR
(per 100000)
Clusters
Russia
72 regions
Proportion of the total change
(%)
1978-79 - 1988-89 – 1993-94 - 1978-79 1969-70 1978-79 1988-89 1969-70
142.4
-196.3
1988-89 1978-79
1993-94 1988-89
Percent of
population
565.9
141.5
-198.7
594.9
100.0
100.0
100.0
100.0
1
35.4
-38.9
103.9
25.0
19.6
17.5
19.4
2
52.9
-62.3
227.5
37.4
31.4
38.2
40.8
3
21.2
-37.5
106.5
15.0
18.9
17.9
17.3
4
27.2
-51.8
118.3
19.2
26.1
19.9
18.1
5
-0.1
-1.1
1.7
-0.1
0.6
0.3
0.3
6
0.1
-1.2
2.1
0.1
0.6
0.3
0.4
7
1.7
-0.5
2.4
1.2
0.2
0.4
1.2
8
2.9
-7.2
6.7
2.0
3.6
1.1
1.6
9
0.6
-0.7
1.2
0.4
0.3
0.2
0.2
10
0.3
-1.8
2.7
0.2
0.9
0.5
0.7
-0.5
4.3
22.0
-0.4
-2.2
3.7
Changes in
the population
distribution
364
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Table 7 examines in detail the contributions by causes of death of each of the 4
main clusters. Major contributions are shown in bold type (black if positive, red if
negative). During the first period, cardiovascular mortality is responsible for the biggest
share of the contribution to the rise in national mortality in all clusters. However, that
contribution differs by cluster. Cluster 2 is obviously the greatest contributor because of
its population size. But, cardiovascular mortality is also the reason why cluster 1's
contribution to the overall mortality increase is greater than that of clusters 3 and 4.
Intriguingly, cluster 3 displays a major counterbalance between “atherosclerotic
cardiosclerosis and ill-defined causes” and both ischaemic heart diseases and
cerebrovascular diseases, reflecting the big improvement in cause of death registration
occurring in that cluster, which contains most of Asian Russia, at the time.
During the second period, “atherosclerotic cardiosclerosis and ill-defined causes”
were responsible for the major contributions by clusters to falling Russian mortality,
reflecting the fact that the anti-alcohol campaign coincided with a general improvement
in cause-of-death registration. This improvement was more significant in the Northern
part of European Russia (cluster 4), which thus contributes more than clusters 1 and 3 to
the total mortality decline notwithstanding the comparable population size of these
clusters.
In the third period, almost every group of causes in all four clusters contributes to
the general increase in total Russian mortality. Barring stomach cancer, which
continues to play a positive role, all causes and clusters deteriorate to an equal extent
with the economic and social crisis.
6. Conclusion
The complex geography of mortality and causes of death, and their time-dependent
variations in a large and heterogeneous country like Russia, make it very difficult to
advance simple, firm conclusions for any single period, and even more so for a global
overview of the four periods analysed here. Arguably, however, at least three
interesting conclusions can be drawn from the present attempt.
Firstly, while there may be no clear and complete coincidence between
geographical variations in total mortality and cause-of-death levels for the territory of
Russia as a whole, there is in European Russia a clear partition between south-west and
north-east, both in terms of total mortality and cause-of-death patterns.
Secondly, when analysing global cause-of-death patterns for all periods combined,
that contrast is clearly confirmed at whole-country level by the prolongation of the
southern part of European Russia by the continuation of the chernoziom belt along the
http://www.demographic-research.org
365
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Table 7:
Contribution of causes of death of each of the 4 main clusters to the
changes in Russian SDR for three periods (percent)
1978-79 - 1969-70
3
1988-89 - 1978-79
1993-94 - 1988-89
1
2
4
1
2
3
4
1
2
3
4
Total
25.0
37.4
15.0
19.2
19.6
31.4
18.9
26.1
17.5
38.2
17.9
19.9
of which, by cause
of death
Infectious
diseases
Stomach cancer
-3.4
-5.3
-2.7
-2.6
2.0
3.4
1.7
1.3
0.2
1.0
0.4
0.2
-3.6
-8.9
-3.1
-4.4
1.2
3.9
1.5
1.6
-0.2
-0.5
-0.2
-0.3
Other digestive
cancers
Cancer of larynx.
trachea. bronchus
and lung
Other cancers
-0.1
-0.2
0.0
0.1
-2.1
-3.2
-1.3
-1.5
0.2
0.3
0.1
0.2
2.4
2.9
1.9
2.3
-2.8
-5.0
-1.9
-2.3
0.4
0.1
0.0
0.2
0.9
0.7
0.5
0.7
-1.1
-2.4
-0.5
-1.0
0.2
0.4
0.1
0.2
Atherosclerotic
cardiosclerosis +
Ill-defined
Other ischaemic
heart diseases
Other heart
diseases
Cerebrovascular
diseases + Other
diseases of the
circulatory system
Influenza+pneumo
nia
Chronic respiratory
diseases
Digestive diseases
5.6
6.3
-10.2
2.9
8.6
16.4
8.9
19.7
4.1
6.7
1.6
4.1
10.7
24.6
15.5
12.2
-5.9
-4.2
-4.3
-7.3
2.4
6.2
3.8
2.1
0.1
-4.1
-0.9
-3.1
0.5
-0.1
0.5
0.7
1.0
2.6
1.2
1.6
7.9
11.2
10.2
6.6
1.2
-3.1
-0.3
0.0
2.1
4.4
2.1
2.9
-2.3
-3.7
-1.3
-1.7
1.4
2.7
1.5
1.0
0.4
1.0
0.5
0.6
-2.5
1.0
-0.6
-2.2
7.8
10.1
3.1
4.6
0.4
0.9
0.6
0.5
Other diseases
Accidental alcohol
poisoning
Suicide +
Homicide+Injury
unspecified
Other external
causes
366
0.8
0.3
0.3
0.0
0.2
0.3
0.0
0.2
0.5
1.3
0.6
0.5
-0.4
-0.2
-0.7
0.6
-0.1
-0.2
-0.1
0.2
0.6
1.6
0.6
0.6
2.0
1.4
1.9
2.7
2.8
3.0
2.9
2.9
1.2
2.2
1.4
1.8
2.9
4.9
3.0
1.8
1.9
2.1
2.5
2.3
2.4
6.2
3.4
2.9
4.0
6.5
1.1
3.4
3.8
7.7
4.6
3.8
1.5
3.6
1.7
1.7
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Kazakhstan border. At the same time, however, a clear split is also made by a second
frontier, that of Ural, between Europe and Asia, for the north-eastern part of Russia.
With the exception of the chernoziom belt, therefore, Siberia appears as a very different
world from European Russia.
Thirdly, looking at relations between the global cause-of-death patterns (Figure 7)
and that for each of the four periods (Figure 3), the period 1988-1989 shows the closest
similarities with the global view, while the three other periods are more peculiar. It is
true that the period 1988-89, which is that of the lowest variance in total mortality, is
also characterized by the highest life expectancy, due to the transient decline in
alcoholism, while cardiovascular diseases had already risen to prominence in the
Russian disease profile. The other three periods have their individual characteristics.
The first remains a transition period in which infectious diseases still play an important
role in the Russian disease profile and above all its geographical variations. In the
second period, mortality is higher and much more tied to cardiovascular diseases and
alcoholism, but geographical differences in total mortality are less important. The
geographical contours become closer to that of the period 1988-89, especially with the
clear split made by Ural. By contrast, the profound economic and social crisis of 199394 makes the geography of cause-of-death patterns of this final period again highly
discrete. So, 1988-89 appears to be a base period, and thus fairly typical of the whole
period. However, that global period marked by such wide fluctuations may itself be a
singular one. Once the 2002 census results become available, it will be interesting to see
whether the geographical pattern of 1988-89 remains a robust benchmark or whether
the more recent trends in Russian life expectancy will result in a new geographical
cause-of-death pattern.
7. Acknowledgements
This paper is the product of collaborative work done under INTAS project n° 1722
“Expectation of life and causes of death in different republics of the ex-USSR: long
term trends and recent changes”.
http://www.demographic-research.org
367
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
References
Andreev, Evgueni M., 1979. – Prodolzhitel'nost' zhizni v SSSR: differentsialniy analiz.
[Length of life in the USSR: a differential analysis], in : Evgueni Andreev and
Anatoli Vishnevski (ed.), Prodolzhitel'nost' zhizni: analiz i modelirovaniye.
[Length of life: analysis and modelling]. – Moskow, Statistika, 136 p.
Andreev, Evgeni and Kvasha Ekaterina, 2002. – Osobennosti pokazateley
mladencheskoy smertnosti v Rossii. [Peculiarity of infant mortality indicators in
Russia], Problemy sotsial'noy gigiyeny, zdravookhraneniya i istorii meditsiny.
[Problems of social hygiene, public health services and history of medicine],
n° 4, p. 15-20.
Gavrilova Natalia S., Evdokushkina Galina N., Semyonova Victoria G., and Gavrilov
Leonid A., 2001. – Economic crises, stress and mortality in Russia. – Chicago,
Center on Aging, University of Chicago, 28 p. (Paper presented at the 2001 PAA
meeting, Washington).
Goskomstat, 2002. – The demographic yearbook of Russia. 2002: Statistical Handbook.
– Moscow, Goskomstat, 397 p.
Jozan Peter and Prokhorskas Remigijus (ed.), 1997. – Atlas of leading and avoidable
causes of death in countries of Central Eastern Europe. – Budapest, Hungarian
CSO Publishing House, Hungarian Central Statistical Office, World Health
Organisation, 324 p.
Leon David A., Chenet Laurent, Shkolnikov Vladimir M., Zakharov Sergei, Shapiro
Judith, Rakhmanova Galina, Vassin Sergei, and McKee Martin, 1997. – Huge
variation in Russian mortality rates 1984-94: artefact, alcohol, or what?, The
Lancet, vol. 350, August 9, p. 383-388.
Meslé France, 2004. – Mortality in Eastern Europe and the former Soviet Union : longterm trends and recent upturns, Demographic Research, p. 46-70. (Special
Collection 2. Determinants of Diverging Trends in Mortality ).
Meslé France and Shkolnikov Vladimir, 2000. – Russie: une crise sanitaire sans
précédents, Espace, Populations, Sociétés, n° 2, p. 253-272.
Meslé France, Shkolnikov Vladimir, Hertrich Véronique, and Vallin Jacques, 1996. –
Tendances récentes de la mortalité par cause en Russie, 1965-1994. – Paris,
INED, 140 p. + 2 floppy disks. (Données statistiques n°2).
Meslé France, Shkolnikov Vladimir, and Vallin Jacques, 1994. – Brusque montée des
morts violentes en Russie, Population, vol. 49, n° 3, p. 780-790.
368
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Meslé France and Vallin Jacques, 2003. – Mortalité et causes de décès en Ukraine au
XXe siècle. – Paris, INED, xvi + 396 p. (Les cahiers de l'INED, cahier n° 152,
with contributions by Vladimir Shkolnikov, Serhii Pyrozhkov and Serguei
Adamets).
Meslé France, Vallin Jacques, Hertrich Véronique, Andreev Evgueni, and Shkolnikov
Vladimir, 2003. – Causes of death in Russia: assessing trends since the 1950s,
in : Irena E. Kotowska and Janina Józwiak (ed.), Population of Central and
Eastern Europe. Challenges and opportunities, p. 389-414. – Warsaw, Statistical
Publishing Establishment, 724 p.
Meslé France, Vallin Jacques, and Shkolnikov Vladimir, 1998. – Reversal of mortality
decline : the case of contemporary Russia, World Health Statistics
Quarterly/Rapport trimestriel de statistiques sanitaires, vol. 51, n° 2-3-4, p. 191206. (“Historical epidemiology: mortality decline, and old and new transitions in
health”, special issue edited by Odile Frank).
Shkolnikov Vladimir M., 1987. – Geograficheskiye faktori prodoljitelnosti jizni.
[Geographical factors of length of life]. Izvestiya AN SSSR, Seriya
Geograficheskaya, vol. 3, n° 12s, p. 35-44. [Academy of Sciences of the SSSR].
Shkolnikov Vladimir, Cornia Giovanni, Leon David, and Meslé France, 1998. – Causes
of the Russian mortality crisis : evidence and interpretations, World
Development, vol. 26, n° 11, p. 1995-2011.
Shkolnikov Vladimir, Meslé France, and Vallin Jacques, 1996. – Health crisis in
Russia. I. Recent trends in life expectancy and causes of death from 1970 to
1993, II. Changes in causes of death: a comparison with France and England and
Wales (1970 to 1993), Population, An English Selection, vol. 8, p. 123-190.
Shkolnikov Vladimir and Nemtsov Alexander, 1997. – The anti-alcohol campaign and
variations in Russian mortality, in: Jose Luis Bobadilla, Christine A. Costello,
and Faith Mitchell (eds), Premature death in the New Independent States,
p. 239-261. – Washington D.C, National Academy Press, 404 p.
Shkolnikov, Vladimir, Valkonen Tapani, and Begun Alexander, 2001. – Measuring
inter-groups inequality in length of life, Genus. Vol. 57, n° 3-4, p. 33-62.
Shkolnikov Vladimir M. and Vassin Sergei A., 1994. – Spatial differences in life
expectancy in European Russia in the 1980s, in : Wolfgang Lutz, Sergei
Scherbov, and Andrei Volkov (éd.), Demographic trends and patterns in the
Soviet Union before 1991, p. 379-402. – London and New York, Routledge,
496 p.
http://www.demographic-research.org
369
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
SPSS, 2002. – SPSS 11.5 Syntax Reference Guide. – Chicago, SPSS Inc.
Vassin Sergei A. and Costello Christine A., 1997. – Spatial, age, and cause-of-death
patterns of mortality in Russia, 1988-1989 , in : José Luis Bodabilla, Christine
A. Costello, and Faith Mitchell (éd.), Premature death in the new independent
states , p. 66-119. – Washington (DC), National Academy Press, 404 p.
Walberg Peder, McKee Martin, Shkolnikov Vladimir M, Chenet Laurent, and Leon
David, 1998. – Economic change, crime and Russian mortality crisis: a regional
analysis, British Medical Journal, vol. 317, August, p. 312-318.
WHO, 1992. – World Health Statistic Annual, 1992. – Geneva, World Health
Organisation, 400 p.
370
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Annex I. Administrative units of the Russian Federation
Table I-1:
List of the 73 administrative units in alphabetical order
Altai Kray & Republic of Altai
71
Lipetsk Oblast
38
Republic of Tatarstan
42
Amur Oblast
96
99
Republic of Tuva
81
Arkhangelsk Oblast
5
Magadan Oblast & Сhukchi
autonomous district
Moscow
20
Rostov Oblast
59
Astrakhan Oblast
43
Moscow Oblast
21
Ryazan Oblast
23
Belgorod Oblast
35
Murmansk Oblast
8
Sakhalin Oblast
100
Bryansk Oblast
15
Nizhny Novgorod Oblast
33
Samara Oblast
46
Chechen & Ingush Republics
56
Novgorod Oblast
12
Saratov Oblast
47
Chelyabinsk Oblast
68
Novosibirsk Oblast
73
Smolensk Oblast
24
Chita Oblast
88
Omsk Oblast
74
St. Petersburg
10
Chuvash Republic
31
Orenburg Oblast
64
Stavropol Kray & Karachaev-
58
Irkutsk Oblast
86
Oryol Oblast
22
Circassian republic
Ivanovo Oblast
17
Penza Oblast
45
Sverdlovsk Oblast
67
Kabardin-Balkar Republic
53
Perm Oblast
65
Tambov Oblast
39
Kaliningrad Oblast
101
Primorsky Kray
94
Tomsk Oblast
75
Kaluga Oblast
18
Pskov Oblast
13
Tula Oblast
26
Kamchatka Oblast
97
Republic of Bashkortostan
61
Tver Oblast
25
Kemerovo Oblast
72
Republic of Buryatia
80
Tyumen Oblast
76
Khabarovsk Kray & Jewish
autonomous oblast
Kirov Oblast
95
Republic of Dagestan
51
Udmurt Republic
62
32
41
Ulyanovsk Oblast
48
Kostroma Oblast
19
R. of Kalmykia - Khalmg
Tangch
Republic of Karelia
3
Vladimir Oblast
16
Krasnodar kray & R. of
Adygeya
Krasnoyarsk Kray & Republic
of Khakasia
Kurgan Oblast
57
Republic of Komi
4
Volgograd Oblast
44
83
Republic of Mari El
29
Vologda Oblast
7
63
Republic of Mordovia
30
Voronezh Oblast
36
Kursk Oblast
37
Republic of North Ossetia
55
Yaroslavl Oblast
27
Leningrad Oblast
11
Republic of Sakha (Yakutia)
91
http://www.demographic-research.org
371
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Table I-2:
List of the 73 administrative units in numerical order
3
Republic of Karelia
32
Kirov Oblast
64
Orenburg Oblast
4
Republic of Komi
33
Nizhny Novgorod Oblast
65
Perm Oblast
5
Arkhangelsk Oblast
35
Belgorod Oblast
67
Sverdlovsk Oblast
7
Vologda Oblast
36
Voronezh Oblast
68
Chelyabinsk Oblast
8
Murmansk Oblast
37
Kursk Oblast
71
Altai Kray & Republic of Altai
10
St. Petersburg
38
Lipetsk Oblast
72
Kemerovo Oblast
11
Leningrad Oblast
39
Tambov Oblast
73
Novosibirsk Oblast
12
Novgorod Oblast
41
74
Omsk Oblast
13
Pskov Oblast
42
Republic of Kalmykia - Khalmg
Tangch
Republic of Tatarstan
75
Tomsk Oblast
15
Bryansk Oblast
43
Astrakhan Oblast
76
Tyumen Oblast
16
Vladimir Oblast
44
Volgograd Oblast
80
Republic of Buryatia
17
Ivanovo Oblast
45
Penza Oblast
81
Republic of Tuva
18
Kaluga Oblast
46
Samara Oblast
83
19
Kostroma Oblast
47
Saratov Oblast
86
Krasnoyarsk Kray & Republic of
Khakasia
Irkutsk Oblast
20
Moscow
48
Ulyanovsk Oblast
88
Chita Oblast
21
Moscow Oblast
51
Republic of Dagestan
91
Republic of Sakha (Yakutia)
22
Oryol Oblast
53
Kabardin-Balkar Republic
94
Primorsky Kray
23
Ryazan Oblast
55
Republic of North Ossetia
95
24
Smolensk Oblast
56
Chechen & Ingush Republics
96
Khabarovsk Kray & Jewish
autonomous oblast
Amur Oblast
25
Tver Oblast
57
97
Kamchatka Oblast
26
Tula Oblast
58
99
100
Magadan Oblast & Сhukchi
autonomous district
Sakhalin Oblast
101
Kaliningrad Oblast
27
Yaroslavl Oblast
59
Krasnodar kray & Republic of
Adygeya
Stavropol Kray & KarachaevCircassian rep.
Rostov Oblast
29
Republic of Mari El
61
Republic of Bashkortostan
30
Republic of Mordovia
62
Udmurt Republic
31
Chuvash Republic
63
Kurgan Oblast
372
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Figure annex:
Map of administrative units used for the geographical analysis.
Numbers are listed above.
99
8
101
Ka li ni ng rad Obl ast
10
11
12
Repu bl ic o f K arel ia
S t . P et ersb urg
13
Murm ansk Oblas t
3
L eni ng a
r d Obl ast
Psk ov Obl ast
Mag ad an Obl ast an d Ch ukc hi A ut on omo us Okrug
5
No vgo rod Obl ast
24
25
7
T ver Obl ast
Smo le nsk Ob la st
V ol og da Ob las t
15 18 21 27
20
19
16 17
26
37 22
33
38 23
32
35
30
39
31 29
36
45
62
48 42
44
59
46
47
61
4
Y aro sl avl Ob la st
Mosc ow
B ryan sk Obl ast
K al ug a Obl ast
Mo sco w Obl as t
Repu bl ic of Ko mi
K amc hat ka Ob la st
Ko str o ma Obl ast
Iv ano vo Obl as t
V la di mir Obl ast
Tu la Ob la st
Oryol Ob la st
91
K ursk Obl as t
Ry aza n Obl ast
76
K ir o v Ob la st
Nizh ny Novg orod Ob la st
Li pet zk Obl as t
83
Be lg orod Ob la st
Re pub li c of Mariy E l
T amb ov Obl as t
Rep ub li c of Mo rdov ia
Chu vas h Repu bl ic
V oron ezh Ob la st
P en za Obl as t
65
P erm Obl ast
Udm urt rep ubl i c
T yume n Obl ast
67
Ulya no vsk Obl ast
Rep ub li c of T at arst an
V ol go grad Ob las t
57
58
53
5556
K ab ardi an-B al kar repu bl ic
Kras noy arsk K a
r y an d Re pu bl ic of Kh aka si a
S ve rdlo vsk Ob las t
Sa rat ov Obl ast
S a mara Obl ast
Ros t ov Obl ast
Kra sno dar kray a nd Rep ubl ic o f A dyg eya
S t avro pol Kray a nd K arac hae v-Cir ca ssi an rep ubl i c
97
Re pu bli c of Sa kha (Y aku t ia)
64
41
Repu bl ic of Ka lmy ki a
Re pub li c of B as hko rt o st an
68
Chel ya bin sk Obl ast
A st a
r kha n Obl ast
43
Ore nbu rg Obl ast
63
K urga n Obl ast
95
75
74
K ha ba ro vsk K ray an d Jewi sh A ut ono mou s Obla st
T oms k Ob la st
Omsk Obla st
Re pub li c of No rth Oss et ia
Che che n an d n
I gu sh Repu bl ics
100
86
73
Sa kha li n Obl ast
96
I rkut sk Obl ast
51
Repu bl ic of Dage st an
No vos ib ir sk Obl a st
72
80
88
A mur Ob la st
Ke merov Ob las t
Ch it a Obl as t
71
Repu bl ic of Bu ryat i a
Al ta i K ray an d Repu bl ic of A lt ai
Rep ubl ic o f T uva
81
94
P rim orsky kra i
St. Petersburg
Moscow
Solid lines are for units used, dotted lines are for autonomous territories (not
distinguished for the analysis)
http://www.demographic-research.org
373
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Annex II. The choice of colour sets for period-specific maps
To make the maps and figures as understandable as possible, we endeavoured to use
colors whose signification would remain near-constant even if the borders of clusters
that they cover varied. The color of each global cluster (for the four periods
simultaneously, Section III) was selected on the basis of the dendrogram produced by
the clustering process, and was chosen from the last step of the process. Two
contrasting colors (blue and red) were attributed to the two large clusters obtained at the
end (Figure 5 above). Then, at every reverse step towards a greater numbers of clusters,
one cluster is subdivided into two. We kept the former color for the larger, and chose a
new but allied color for the smaller. Finally, we selected 10 colours for the global
clusters.
When repeating the analysis separately for each period, the geographic contents of
clusters clearly vary. But most share many characteristics with clusters of the global
analysis, and we aimed to use the same colors to reflect such similarities. To do that, we
calculated the distance between period-specific clusters and global clusters in terms of
cause-of-death levels and patterns for each period. We first computed the Euclidian
distance between each region at one period to global clusters:
(
)
1
2 2

 ∑ (W (r , t , i ) − W (Rk , i )) ⋅ SDR(0, t , i )  ,
 i

where Rk is the set of the nk regions included in cluster k and
W ( Rk , i ) =
1
4 nk
∑ ∑ W (r , t , i ) ,
t
r ∈ Rk
and, finally, the distance between each period-specific cluster and the global clusters
was obtained as the average of the Euclidian distances of regions belonging to the
period-specific cluster.
On the basis of these distances, we assigned the color of the nearest global cluster
to each period-specific cluster (Table II-1). Note that a period-specific cluster and the
nearest global cluster are not necessarily the closest geographically. We also used the
same colors to plot the stars representing the cause-of-death patterns (Figures 4, 6,
and 9).
In 3 of 4 periods, the same global cluster was the nearest for two period-specific
clusters. We addressed this problem by adding a specific pattern to the selected color to
maintain the differentiation between the two period-specific clusters. For example, in
1978-1979 the global cluster 1 (red) is the nearest for period-specific clusters 1 and 8,
but at this period cluster 8 includes only the Republic of Dagestan, and we attributed the
374
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
red color to cluster 1, shadowed by a same-colour pattern for cluster 8, which will be
numbered 11, in that case. Conversely, the color of global cluster 8 (orange) is not used
for any period-specific cluster, since that global cluster never appeared as the nearest
for any period-specific cluster.
Table II-2 gives the number of units for each cluster (either global or period
specific), including a column showing the color used for each on the maps. Note that
lines 1 and 6 also show the alternative pattern used to identify the additional cluster 11.
Table II-1:
Period-specific
cluster number
Distance between period-specific clusters and global clusters
Global cluster number
1
2
3
4
5
6
7
8
9
10
1970
1
0.0005
0.0080
0.0197
0.0130
0.0737
0.0850
0.0591
0.0151
0.0757
0.0383
2
0.0123
0.0004
0.0078
0.0089
0.0553
0.0563
0.0652
0.0297
0.0633
0.0378
3
0.0136
0.0074
0.0031
0.0059
0.0371
0.0465
0.0848
0.0237
0.0385
0.0283
4
0.0129
0.0120
0.0144
0.0012
0.0464
0.0606
0.1114
0.0222
0.0737
0.0473
5
0.1117
0.1187
0.1097
0.0811
0.0538
0.1049
0.2808
0.1350
0.1454
0.1180
6
0.1155
0.1039
0.0696
0.0797
0.0483
0.0313
0.2278
0.1034
0.0983
0.1141
7
0.0555
0.0463
0.0633
0.0859
0.1382
0.1367
0.0101
0.0806
0.0996
0.0683
8
0.0092
0.0310
0.0446
0.0288
0.0966
0.1150
0.0858
0.0158
0.1056
0.0611
9
0.0735
0.0615
0.0399
0.0661
0.0758
0.0827
0.0948
0.0682
0.0165
0.0438
10
0.0780
0.0657
0.0501
0.0707
0.0818
0.0820
0.1319
0.0956
0.0254
0.0139
1
0.0002
0.0093
0.0199
0.0122
0.0715
0.0841
0.0645
0.0142
0.0777
0.0406
2
0.0099
0.0001
0.0085
0.0088
0.0571
0.0594
0.0620
0.0277
0.0653
0.0368
3
0.0248
0.0109
0.0003
0.0105
0.0357
0.0338
0.0885
0.0290
0.0331
0.0365
4
0.0128
0.0111
0.0138
0.0008
0.0424
0.0576
0.1119
0.0239
0.0732
0.0461
5
0.0718
0.0554
0.0384
0.0541
0.0154
0.0213
0.1406
0.0966
0.0688
0.0597
6
0.0699
0.0549
0.0331
0.0406
0.0277
0.0128
0.1657
0.0667
0.0561
0.0720
7
0.0866
0.0810
0.0984
0.1248
0.1909
0.1842
0.0048
0.1046
0.1366
0.1066
8
0.0148
0.0413
0.0525
0.0333
0.0969
0.1230
0.1040
0.0187
0.1140
0.0704
1979
9
0.1322
0.1037
0.0608
0.0994
0.0714
0.0589
0.1902
0.1216
0.0131
0.0736
10
0.0448
0.0470
0.0437
0.0506
0.0634
0.0797
0.1067
0.0732
0.0443
0.0029
http://www.demographic-research.org
375
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Table II-1:
Period-specific
cluster number
(Continued)
Global cluster number
1
2
3
4
5
6
7
8
9
10
1989
1
0.0003
0.0102
0.0218
0.0133
0.0689
0.0843
0.0661
0.0196
0.0803
0.0388
2
0.0070
0.0002
0.0088
0.0068
0.0556
0.0616
0.0653
0.0244
0.0651
0.0359
3
0.0236
0.0092
0.0006
0.0110
0.0402
0.0338
0.0795
0.0278
0.0382
0.0382
4
0.0144
0.0056
0.0054
0.0016
0.0413
0.0449
0.0933
0.0235
0.0596
0.0423
5
0.1090
0.0731
0.0505
0.0622
0.0304
0.0230
0.2009
0.1129
0.0938
0.0990
6
0.1164
0.0806
0.0599
0.0773
0.0391
0.0154
0.2022
0.1406
0.0772
0.0813
7
0.0639
0.0710
0.0898
0.1087
0.1787
0.1760
0.0037
0.0840
0.1403
0.1005
8
0.0429
0.0413
0.0316
0.0367
0.1021
0.0903
0.0946
0.0091
0.0794
0.0848
9
0.0633
0.0584
0.0330
0.0568
0.0629
0.0646
0.1323
0.0700
0.0097
0.0319
10
0.0270
0.0254
0.0269
0.0320
0.0501
0.0587
0.0842
0.0527
0.0497
0.0040
1
0.0006
0.0093
0.0241
0.0132
0.0738
0.0875
0.0674
0.0218
0.0885
0.0437
2
0.0070
0.0002
0.0103
0.0084
0.0592
0.0649
0.0620
0.0260
0.0669
0.0359
3
0.0271
0.0122
0.0012
0.0139
0.0348
0.0295
0.0824
0.0331
0.0348
0.0377
4
0.0142
0.0075
0.0066
0.0006
0.0367
0.0439
0.1008
0.0232
0.0602
0.0414
5
0.1202
0.1057
0.0705
0.0918
0.0268
0.0530
0.1992
0.1371
0.0568
0.0880
6
0.1234
0.0879
0.0618
0.0912
0.0465
0.0264
0.1685
0.1406
0.0825
0.0945
7
0.0757
0.0818
0.1048
0.1202
0.2135
0.2077
0.0077
0.0871
0.1687
0.1241
8
0.0730
0.0639
0.0503
0.0605
0.1302
0.1077
0.1161
0.0272
0.0973
0.1126
1994
9
0.1020
0.0866
0.0560
0.0908
0.0788
0.0725
0.1521
0.1164
0.0113
0.0545
10
0.0383
0.0359
0.0432
0.0480
0.0713
0.0841
0.0787
0.0650
0.0622
0.0076
376
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Table II-2:
Number of administrative units in the global clusters and in
corresponding nearest cluster(s) for specific periods
Global cluster
Number of units in the nearest cluster(s)
Number
of units
1969-70
1978-79
1988-89
1993-94
19
10/1
25/1
13
18
2
18
39
29
18
6
3
11
10
8
19
20
4
17
5
1
16
19
5
1
1
4
-
1
1
1
2
1/2
3
7
1
4
1
1
1
8
2
-
-
1
1
9
1
1
1
1
2
10
1
1
1
1
1
No data
1
Total
73
Number
1
6
Color
/
/
http://www.demographic-research.org
1
73
73
73
73
377
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
Annex III. Why use
SDR as a weight?
Suppose there are two causes with very similar variations across regions. A desirable
property of a distance measure would be that such measure does not change if we
aggregate the two causes of death into one.
Let SDR (r , t ,2 ) = λ ⋅ SDR(r , t ,1) + С . Let 1&2 denote the sum of causes 1 and
2. Then SDR (r , t ,1 & 2 ) = (1 + λ ) ⋅ SDR (r , t ,1) + С and for any pair of regions r1
and r2:
(W (r2 , t ,1) − W (r1 , t ,1)) = (W (r2 , t , 2) − W (r1 , t , 2)) = (W (r2 , t ,1 & 2) − W (r1 , t ,1 & 2))
It is clear that the sum of simple Euclidian distances D[W(r2,t,1),W(r1,t,1)] and
D[W(r2,t,2),W(r1,t,1)] will be higher than the distance for the summary cause 1&2
D[W(r2,t,1&2),W(r1,t,1&2)]. However with weights, equal to
SDR (0, t , i ) , the two
distances are equal:
(W (r2 , t ,1) − W (r1 , t ,1)) 2 SDR(0, t ,1) + (W (r2 , t , 2) − W (r1 , t , 2)) 2 SDR(0, t ,2) =
(W (r2 , t ,1 & 2) − W (r1 , t ,1 & 2)) 2 SDR(0, t ,1 & 2)
378
http://www.demographic-research.org
Demographic Research: Volume 12, Article 13
Annex IV. Life expectancy at birth for regions in each SDR interval
(see footnote 10)
Males
1969-70
1978-79
1988-89
1993-94
SDR intervals
<12.22
69.80
12.22-15.66
66.82
66.22
67.42
15.66 to 19.10
64.43
64.43
64.96
65.25
19.10 to 22.54
61.98
61.55
63.50
61.27
22.54 to 25.98
60.56
59.25
62.94
25.98 to 29.42
58.60
57.90
>= 29.42
56.44
Females
1969-70
1978-79
<6.92
82.62
81.16
6.92-8.62
77.08
75.43
76.68
8.62 to 10.32
74.68
74.86
75.29
74.74
10.32 to 12.02
73.27
73.23
74.12
73.12
12.02 to 13.72
70.99
71.15
72.55
71.66
13.72 to 15.43
70.26
68.18
70.20
69.59
68.21
68.90
58.16
58.87
56.45
48.70
1988-89
1993-94
SDR intervals
>= 15.43
http://www.demographic-research.org
379
Vallin et al: Geographical diversity of cause-of-death patterns and trends in Russia
380
http://www.demographic-research.org